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 #include <net/rstreason.h>
33
34 #include "dccp.h"
35 #include "ipv6.h"
36 #include "feat.h"
37
38 struct dccp_v6_pernet {
39 struct sock *v6_ctl_sk;
40 };
41
42 static unsigned int dccp_v6_pernet_id __read_mostly;
43
44 /* The per-net v6_ctl_sk is used for sending RSTs and ACKs */
45
46 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
47 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
48
49 /* 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)50 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
51 const struct in6_addr *saddr,
52 const struct in6_addr *daddr)
53 {
54 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
55 }
56
dccp_v6_send_check(struct sock * sk,struct sk_buff * skb)57 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
58 {
59 struct ipv6_pinfo *np = inet6_sk(sk);
60 struct dccp_hdr *dh = dccp_hdr(skb);
61
62 dccp_csum_outgoing(skb);
63 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
64 }
65
dccp_v6_init_sequence(struct sk_buff * skb)66 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
67 {
68 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
69 ipv6_hdr(skb)->saddr.s6_addr32,
70 dccp_hdr(skb)->dccph_dport,
71 dccp_hdr(skb)->dccph_sport );
72
73 }
74
dccp_v6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)75 static int dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
76 u8 type, u8 code, int offset, __be32 info)
77 {
78 const struct ipv6hdr *hdr;
79 const struct dccp_hdr *dh;
80 struct dccp_sock *dp;
81 struct ipv6_pinfo *np;
82 struct sock *sk;
83 int err;
84 __u64 seq;
85 struct net *net = dev_net(skb->dev);
86
87 if (!pskb_may_pull(skb, offset + sizeof(*dh)))
88 return -EINVAL;
89 dh = (struct dccp_hdr *)(skb->data + offset);
90 if (!pskb_may_pull(skb, offset + __dccp_basic_hdr_len(dh)))
91 return -EINVAL;
92 hdr = (const struct ipv6hdr *)skb->data;
93 dh = (struct dccp_hdr *)(skb->data + offset);
94
95 sk = __inet6_lookup_established(net, &dccp_hashinfo,
96 &hdr->daddr, dh->dccph_dport,
97 &hdr->saddr, ntohs(dh->dccph_sport),
98 inet6_iif(skb), 0);
99
100 if (!sk) {
101 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
102 ICMP6_MIB_INERRORS);
103 return -ENOENT;
104 }
105
106 if (sk->sk_state == DCCP_TIME_WAIT) {
107 inet_twsk_put(inet_twsk(sk));
108 return 0;
109 }
110 seq = dccp_hdr_seq(dh);
111 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
112 dccp_req_err(sk, seq);
113 return 0;
114 }
115
116 bh_lock_sock(sk);
117 if (sock_owned_by_user(sk))
118 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
119
120 if (sk->sk_state == DCCP_CLOSED)
121 goto out;
122
123 dp = dccp_sk(sk);
124 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
125 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
126 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
127 goto out;
128 }
129
130 np = inet6_sk(sk);
131
132 if (type == NDISC_REDIRECT) {
133 if (!sock_owned_by_user(sk)) {
134 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
135
136 if (dst)
137 dst->ops->redirect(dst, sk, skb);
138 }
139 goto out;
140 }
141
142 if (type == ICMPV6_PKT_TOOBIG) {
143 struct dst_entry *dst = NULL;
144
145 if (!ip6_sk_accept_pmtu(sk))
146 goto out;
147
148 if (sock_owned_by_user(sk))
149 goto out;
150 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
151 goto out;
152
153 dst = inet6_csk_update_pmtu(sk, ntohl(info));
154 if (!dst)
155 goto out;
156
157 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
158 dccp_sync_mss(sk, dst_mtu(dst));
159 goto out;
160 }
161
162 icmpv6_err_convert(type, code, &err);
163
164 /* Might be for an request_sock */
165 switch (sk->sk_state) {
166 case DCCP_REQUESTING:
167 case DCCP_RESPOND: /* Cannot happen.
168 It can, it SYNs are crossed. --ANK */
169 if (!sock_owned_by_user(sk)) {
170 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
171 sk->sk_err = err;
172 /*
173 * Wake people up to see the error
174 * (see connect in sock.c)
175 */
176 sk_error_report(sk);
177 dccp_done(sk);
178 } else {
179 WRITE_ONCE(sk->sk_err_soft, err);
180 }
181 goto out;
182 }
183
184 if (!sock_owned_by_user(sk) && inet6_test_bit(RECVERR6, sk)) {
185 sk->sk_err = err;
186 sk_error_report(sk);
187 } else {
188 WRITE_ONCE(sk->sk_err_soft, err);
189 }
190 out:
191 bh_unlock_sock(sk);
192 sock_put(sk);
193 return 0;
194 }
195
196
dccp_v6_send_response(const struct sock * sk,struct request_sock * req)197 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
198 {
199 struct inet_request_sock *ireq = inet_rsk(req);
200 struct ipv6_pinfo *np = inet6_sk(sk);
201 struct sk_buff *skb;
202 struct in6_addr *final_p, final;
203 struct flowi6 fl6;
204 int err = -1;
205 struct dst_entry *dst;
206
207 memset(&fl6, 0, sizeof(fl6));
208 fl6.flowi6_proto = IPPROTO_DCCP;
209 fl6.daddr = ireq->ir_v6_rmt_addr;
210 fl6.saddr = ireq->ir_v6_loc_addr;
211 fl6.flowlabel = 0;
212 fl6.flowi6_oif = ireq->ir_iif;
213 fl6.fl6_dport = ireq->ir_rmt_port;
214 fl6.fl6_sport = htons(ireq->ir_num);
215 security_req_classify_flow(req, flowi6_to_flowi_common(&fl6));
216
217
218 rcu_read_lock();
219 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
220 rcu_read_unlock();
221
222 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
223 if (IS_ERR(dst)) {
224 err = PTR_ERR(dst);
225 dst = NULL;
226 goto done;
227 }
228
229 skb = dccp_make_response(sk, dst, req);
230 if (skb != NULL) {
231 struct dccp_hdr *dh = dccp_hdr(skb);
232 struct ipv6_txoptions *opt;
233
234 dh->dccph_checksum = dccp_v6_csum_finish(skb,
235 &ireq->ir_v6_loc_addr,
236 &ireq->ir_v6_rmt_addr);
237 fl6.daddr = ireq->ir_v6_rmt_addr;
238 rcu_read_lock();
239 opt = ireq->ipv6_opt;
240 if (!opt)
241 opt = rcu_dereference(np->opt);
242 err = ip6_xmit(sk, skb, &fl6, READ_ONCE(sk->sk_mark), opt,
243 np->tclass, READ_ONCE(sk->sk_priority));
244 rcu_read_unlock();
245 err = net_xmit_eval(err);
246 }
247
248 done:
249 dst_release(dst);
250 return err;
251 }
252
dccp_v6_reqsk_destructor(struct request_sock * req)253 static void dccp_v6_reqsk_destructor(struct request_sock *req)
254 {
255 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
256 kfree(inet_rsk(req)->ipv6_opt);
257 kfree_skb(inet_rsk(req)->pktopts);
258 }
259
dccp_v6_ctl_send_reset(const struct sock * sk,struct sk_buff * rxskb,enum sk_rst_reason reason)260 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb,
261 enum sk_rst_reason reason)
262 {
263 const struct ipv6hdr *rxip6h;
264 struct sk_buff *skb;
265 struct flowi6 fl6;
266 struct net *net = dev_net(skb_dst(rxskb)->dev);
267 struct dccp_v6_pernet *pn;
268 struct sock *ctl_sk;
269 struct dst_entry *dst;
270
271 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
272 return;
273
274 if (!ipv6_unicast_destination(rxskb))
275 return;
276
277 pn = net_generic(net, dccp_v6_pernet_id);
278 ctl_sk = pn->v6_ctl_sk;
279 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
280 if (skb == NULL)
281 return;
282
283 rxip6h = ipv6_hdr(rxskb);
284 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
285 &rxip6h->daddr);
286
287 memset(&fl6, 0, sizeof(fl6));
288 fl6.daddr = rxip6h->saddr;
289 fl6.saddr = rxip6h->daddr;
290
291 fl6.flowi6_proto = IPPROTO_DCCP;
292 fl6.flowi6_oif = inet6_iif(rxskb);
293 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
294 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
295 security_skb_classify_flow(rxskb, flowi6_to_flowi_common(&fl6));
296
297 /* sk = NULL, but it is safe for now. RST socket required. */
298 dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL);
299 if (!IS_ERR(dst)) {
300 skb_dst_set(skb, dst);
301 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0, 0);
302 DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
303 DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
304 return;
305 }
306
307 kfree_skb(skb);
308 }
309
310 static struct request_sock_ops dccp6_request_sock_ops = {
311 .family = AF_INET6,
312 .obj_size = sizeof(struct dccp6_request_sock),
313 .rtx_syn_ack = dccp_v6_send_response,
314 .send_ack = dccp_reqsk_send_ack,
315 .destructor = dccp_v6_reqsk_destructor,
316 .send_reset = dccp_v6_ctl_send_reset,
317 .syn_ack_timeout = dccp_syn_ack_timeout,
318 };
319
dccp_v6_conn_request(struct sock * sk,struct sk_buff * skb)320 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
321 {
322 struct request_sock *req;
323 struct dccp_request_sock *dreq;
324 struct inet_request_sock *ireq;
325 struct ipv6_pinfo *np = inet6_sk(sk);
326 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
327 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
328
329 if (skb->protocol == htons(ETH_P_IP))
330 return dccp_v4_conn_request(sk, skb);
331
332 if (!ipv6_unicast_destination(skb))
333 return 0; /* discard, don't send a reset here */
334
335 if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) {
336 __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS);
337 return 0;
338 }
339
340 if (dccp_bad_service_code(sk, service)) {
341 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
342 goto drop;
343 }
344 /*
345 * There are no SYN attacks on IPv6, yet...
346 */
347 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
348 if (inet_csk_reqsk_queue_is_full(sk))
349 goto drop;
350
351 if (sk_acceptq_is_full(sk))
352 goto drop;
353
354 req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
355 if (req == NULL)
356 goto drop;
357
358 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
359 goto drop_and_free;
360
361 dreq = dccp_rsk(req);
362 if (dccp_parse_options(sk, dreq, skb))
363 goto drop_and_free;
364
365 ireq = inet_rsk(req);
366 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
367 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
368 ireq->ireq_family = AF_INET6;
369 ireq->ir_mark = inet_request_mark(sk, skb);
370
371 if (security_inet_conn_request(sk, skb, req))
372 goto drop_and_free;
373
374 if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
375 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
376 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
377 refcount_inc(&skb->users);
378 ireq->pktopts = skb;
379 }
380 ireq->ir_iif = READ_ONCE(sk->sk_bound_dev_if);
381
382 /* So that link locals have meaning */
383 if (!ireq->ir_iif &&
384 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
385 ireq->ir_iif = inet6_iif(skb);
386
387 /*
388 * Step 3: Process LISTEN state
389 *
390 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
391 *
392 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
393 */
394 dreq->dreq_isr = dcb->dccpd_seq;
395 dreq->dreq_gsr = dreq->dreq_isr;
396 dreq->dreq_iss = dccp_v6_init_sequence(skb);
397 dreq->dreq_gss = dreq->dreq_iss;
398 dreq->dreq_service = service;
399
400 if (dccp_v6_send_response(sk, req))
401 goto drop_and_free;
402
403 if (unlikely(!inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT)))
404 reqsk_free(req);
405 else
406 reqsk_put(req);
407
408 return 0;
409
410 drop_and_free:
411 reqsk_free(req);
412 drop:
413 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
414 return -1;
415 }
416
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)417 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
418 struct sk_buff *skb,
419 struct request_sock *req,
420 struct dst_entry *dst,
421 struct request_sock *req_unhash,
422 bool *own_req)
423 {
424 struct inet_request_sock *ireq = inet_rsk(req);
425 struct ipv6_pinfo *newnp;
426 const struct ipv6_pinfo *np = inet6_sk(sk);
427 struct ipv6_txoptions *opt;
428 struct inet_sock *newinet;
429 struct dccp6_sock *newdp6;
430 struct sock *newsk;
431
432 if (skb->protocol == htons(ETH_P_IP)) {
433 /*
434 * v6 mapped
435 */
436 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
437 req_unhash, own_req);
438 if (newsk == NULL)
439 return NULL;
440
441 newdp6 = (struct dccp6_sock *)newsk;
442 newinet = inet_sk(newsk);
443 newinet->pinet6 = &newdp6->inet6;
444 newnp = inet6_sk(newsk);
445
446 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
447
448 newnp->saddr = newsk->sk_v6_rcv_saddr;
449
450 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
451 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
452 newnp->pktoptions = NULL;
453 newnp->opt = NULL;
454 newnp->ipv6_mc_list = NULL;
455 newnp->ipv6_ac_list = NULL;
456 newnp->ipv6_fl_list = NULL;
457 newnp->mcast_oif = inet_iif(skb);
458 newnp->mcast_hops = ip_hdr(skb)->ttl;
459
460 /*
461 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
462 * here, dccp_create_openreq_child now does this for us, see the comment in
463 * that function for the gory details. -acme
464 */
465
466 /* It is tricky place. Until this moment IPv4 tcp
467 worked with IPv6 icsk.icsk_af_ops.
468 Sync it now.
469 */
470 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
471
472 return newsk;
473 }
474
475
476 if (sk_acceptq_is_full(sk))
477 goto out_overflow;
478
479 if (!dst) {
480 struct flowi6 fl6;
481
482 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
483 if (!dst)
484 goto out;
485 }
486
487 newsk = dccp_create_openreq_child(sk, req, skb);
488 if (newsk == NULL)
489 goto out_nonewsk;
490
491 /*
492 * No need to charge this sock to the relevant IPv6 refcnt debug socks
493 * count here, dccp_create_openreq_child now does this for us, see the
494 * comment in that function for the gory details. -acme
495 */
496
497 ip6_dst_store(newsk, dst, NULL, NULL);
498 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
499 NETIF_F_TSO);
500 newdp6 = (struct dccp6_sock *)newsk;
501 newinet = inet_sk(newsk);
502 newinet->pinet6 = &newdp6->inet6;
503 newnp = inet6_sk(newsk);
504
505 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
506
507 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
508 newnp->saddr = ireq->ir_v6_loc_addr;
509 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
510 newsk->sk_bound_dev_if = ireq->ir_iif;
511
512 /* Now IPv6 options...
513
514 First: no IPv4 options.
515 */
516 newinet->inet_opt = NULL;
517
518 /* Clone RX bits */
519 newnp->rxopt.all = np->rxopt.all;
520
521 newnp->ipv6_mc_list = NULL;
522 newnp->ipv6_ac_list = NULL;
523 newnp->ipv6_fl_list = NULL;
524 newnp->pktoptions = NULL;
525 newnp->opt = NULL;
526 newnp->mcast_oif = inet6_iif(skb);
527 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
528
529 /*
530 * Clone native IPv6 options from listening socket (if any)
531 *
532 * Yes, keeping reference count would be much more clever, but we make
533 * one more one thing there: reattach optmem to newsk.
534 */
535 opt = ireq->ipv6_opt;
536 if (!opt)
537 opt = rcu_dereference(np->opt);
538 if (opt) {
539 opt = ipv6_dup_options(newsk, opt);
540 RCU_INIT_POINTER(newnp->opt, opt);
541 }
542 inet_csk(newsk)->icsk_ext_hdr_len = 0;
543 if (opt)
544 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
545 opt->opt_flen;
546
547 dccp_sync_mss(newsk, dst_mtu(dst));
548
549 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
550 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
551
552 if (__inet_inherit_port(sk, newsk) < 0) {
553 inet_csk_prepare_forced_close(newsk);
554 dccp_done(newsk);
555 goto out;
556 }
557 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
558 /* Clone pktoptions received with SYN, if we own the req */
559 if (*own_req && ireq->pktopts) {
560 newnp->pktoptions = skb_clone_and_charge_r(ireq->pktopts, newsk);
561 consume_skb(ireq->pktopts);
562 ireq->pktopts = NULL;
563 }
564
565 return newsk;
566
567 out_overflow:
568 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
569 out_nonewsk:
570 dst_release(dst);
571 out:
572 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
573 return NULL;
574 }
575
576 /* The socket must have it's spinlock held when we get
577 * here.
578 *
579 * We have a potential double-lock case here, so even when
580 * doing backlog processing we use the BH locking scheme.
581 * This is because we cannot sleep with the original spinlock
582 * held.
583 */
dccp_v6_do_rcv(struct sock * sk,struct sk_buff * skb)584 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
585 {
586 struct ipv6_pinfo *np = inet6_sk(sk);
587 struct sk_buff *opt_skb = NULL;
588
589 /* Imagine: socket is IPv6. IPv4 packet arrives,
590 goes to IPv4 receive handler and backlogged.
591 From backlog it always goes here. Kerboom...
592 Fortunately, dccp_rcv_established and rcv_established
593 handle them correctly, but it is not case with
594 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
595 */
596
597 if (skb->protocol == htons(ETH_P_IP))
598 return dccp_v4_do_rcv(sk, skb);
599
600 if (sk_filter(sk, skb))
601 goto discard;
602
603 /*
604 * socket locking is here for SMP purposes as backlog rcv is currently
605 * called with bh processing disabled.
606 */
607
608 /* Do Stevens' IPV6_PKTOPTIONS.
609
610 Yes, guys, it is the only place in our code, where we
611 may make it not affecting IPv4.
612 The rest of code is protocol independent,
613 and I do not like idea to uglify IPv4.
614
615 Actually, all the idea behind IPV6_PKTOPTIONS
616 looks not very well thought. For now we latch
617 options, received in the last packet, enqueued
618 by tcp. Feel free to propose better solution.
619 --ANK (980728)
620 */
621 if (np->rxopt.all && sk->sk_state != DCCP_LISTEN)
622 opt_skb = skb_clone_and_charge_r(skb, sk);
623
624 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
625 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
626 goto reset;
627 if (opt_skb)
628 goto ipv6_pktoptions;
629 return 0;
630 }
631
632 /*
633 * Step 3: Process LISTEN state
634 * If S.state == LISTEN,
635 * If P.type == Request or P contains a valid Init Cookie option,
636 * (* Must scan the packet's options to check for Init
637 * Cookies. Only Init Cookies are processed here,
638 * however; other options are processed in Step 8. This
639 * scan need only be performed if the endpoint uses Init
640 * Cookies *)
641 * (* Generate a new socket and switch to that socket *)
642 * Set S := new socket for this port pair
643 * S.state = RESPOND
644 * Choose S.ISS (initial seqno) or set from Init Cookies
645 * Initialize S.GAR := S.ISS
646 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
647 * Continue with S.state == RESPOND
648 * (* A Response packet will be generated in Step 11 *)
649 * Otherwise,
650 * Generate Reset(No Connection) unless P.type == Reset
651 * Drop packet and return
652 *
653 * NOTE: the check for the packet types is done in
654 * dccp_rcv_state_process
655 */
656
657 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
658 goto reset;
659 if (opt_skb)
660 goto ipv6_pktoptions;
661 return 0;
662
663 reset:
664 dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
665 discard:
666 if (opt_skb != NULL)
667 __kfree_skb(opt_skb);
668 kfree_skb(skb);
669 return 0;
670
671 /* Handling IPV6_PKTOPTIONS skb the similar
672 * way it's done for net/ipv6/tcp_ipv6.c
673 */
674 ipv6_pktoptions:
675 if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
676 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
677 WRITE_ONCE(np->mcast_oif, inet6_iif(opt_skb));
678 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
679 WRITE_ONCE(np->mcast_hops, ipv6_hdr(opt_skb)->hop_limit);
680 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
681 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
682 if (inet6_test_bit(REPFLOW, sk))
683 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
684 if (ipv6_opt_accepted(sk, opt_skb,
685 &DCCP_SKB_CB(opt_skb)->header.h6)) {
686 memmove(IP6CB(opt_skb),
687 &DCCP_SKB_CB(opt_skb)->header.h6,
688 sizeof(struct inet6_skb_parm));
689 opt_skb = xchg(&np->pktoptions, opt_skb);
690 } else {
691 __kfree_skb(opt_skb);
692 opt_skb = xchg(&np->pktoptions, NULL);
693 }
694 }
695
696 kfree_skb(opt_skb);
697 return 0;
698 }
699
dccp_v6_rcv(struct sk_buff * skb)700 static int dccp_v6_rcv(struct sk_buff *skb)
701 {
702 const struct dccp_hdr *dh;
703 bool refcounted;
704 struct sock *sk;
705 int min_cov;
706
707 /* Step 1: Check header basics */
708
709 if (dccp_invalid_packet(skb))
710 goto discard_it;
711
712 /* Step 1: If header checksum is incorrect, drop packet and return. */
713 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
714 &ipv6_hdr(skb)->daddr)) {
715 DCCP_WARN("dropped packet with invalid checksum\n");
716 goto discard_it;
717 }
718
719 dh = dccp_hdr(skb);
720
721 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
722 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
723
724 if (dccp_packet_without_ack(skb))
725 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
726 else
727 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
728
729 lookup:
730 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
731 dh->dccph_sport, dh->dccph_dport,
732 inet6_iif(skb), 0, &refcounted);
733 if (!sk) {
734 dccp_pr_debug("failed to look up flow ID in table and "
735 "get corresponding socket\n");
736 goto no_dccp_socket;
737 }
738
739 /*
740 * Step 2:
741 * ... or S.state == TIMEWAIT,
742 * Generate Reset(No Connection) unless P.type == Reset
743 * Drop packet and return
744 */
745 if (sk->sk_state == DCCP_TIME_WAIT) {
746 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
747 inet_twsk_put(inet_twsk(sk));
748 goto no_dccp_socket;
749 }
750
751 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
752 struct request_sock *req = inet_reqsk(sk);
753 struct sock *nsk;
754
755 sk = req->rsk_listener;
756 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
757 inet_csk_reqsk_queue_drop_and_put(sk, req);
758 goto lookup;
759 }
760 sock_hold(sk);
761 refcounted = true;
762 nsk = dccp_check_req(sk, skb, req);
763 if (!nsk) {
764 reqsk_put(req);
765 goto discard_and_relse;
766 }
767 if (nsk == sk) {
768 reqsk_put(req);
769 } else if (dccp_child_process(sk, nsk, skb)) {
770 dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
771 goto discard_and_relse;
772 } else {
773 sock_put(sk);
774 return 0;
775 }
776 }
777 /*
778 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
779 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
780 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
781 */
782 min_cov = dccp_sk(sk)->dccps_pcrlen;
783 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
784 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
785 dh->dccph_cscov, min_cov);
786 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
787 goto discard_and_relse;
788 }
789
790 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
791 goto discard_and_relse;
792 nf_reset_ct(skb);
793
794 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
795 refcounted) ? -1 : 0;
796
797 no_dccp_socket:
798 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
799 goto discard_it;
800 /*
801 * Step 2:
802 * If no socket ...
803 * Generate Reset(No Connection) unless P.type == Reset
804 * Drop packet and return
805 */
806 if (dh->dccph_type != DCCP_PKT_RESET) {
807 DCCP_SKB_CB(skb)->dccpd_reset_code =
808 DCCP_RESET_CODE_NO_CONNECTION;
809 dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
810 }
811
812 discard_it:
813 kfree_skb(skb);
814 return 0;
815
816 discard_and_relse:
817 if (refcounted)
818 sock_put(sk);
819 goto discard_it;
820 }
821
dccp_v6_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)822 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
823 int addr_len)
824 {
825 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
826 struct inet_connection_sock *icsk = inet_csk(sk);
827 struct inet_sock *inet = inet_sk(sk);
828 struct ipv6_pinfo *np = inet6_sk(sk);
829 struct dccp_sock *dp = dccp_sk(sk);
830 struct in6_addr *saddr = NULL, *final_p, final;
831 struct ipv6_txoptions *opt;
832 struct flowi6 fl6;
833 struct dst_entry *dst;
834 int addr_type;
835 int err;
836
837 dp->dccps_role = DCCP_ROLE_CLIENT;
838
839 if (addr_len < SIN6_LEN_RFC2133)
840 return -EINVAL;
841
842 if (usin->sin6_family != AF_INET6)
843 return -EAFNOSUPPORT;
844
845 memset(&fl6, 0, sizeof(fl6));
846
847 if (inet6_test_bit(SNDFLOW, sk)) {
848 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
849 IP6_ECN_flow_init(fl6.flowlabel);
850 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
851 struct ip6_flowlabel *flowlabel;
852 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
853 if (IS_ERR(flowlabel))
854 return -EINVAL;
855 fl6_sock_release(flowlabel);
856 }
857 }
858 /*
859 * connect() to INADDR_ANY means loopback (BSD'ism).
860 */
861 if (ipv6_addr_any(&usin->sin6_addr))
862 usin->sin6_addr.s6_addr[15] = 1;
863
864 addr_type = ipv6_addr_type(&usin->sin6_addr);
865
866 if (addr_type & IPV6_ADDR_MULTICAST)
867 return -ENETUNREACH;
868
869 if (addr_type & IPV6_ADDR_LINKLOCAL) {
870 if (addr_len >= sizeof(struct sockaddr_in6) &&
871 usin->sin6_scope_id) {
872 /* If interface is set while binding, indices
873 * must coincide.
874 */
875 if (sk->sk_bound_dev_if &&
876 sk->sk_bound_dev_if != usin->sin6_scope_id)
877 return -EINVAL;
878
879 sk->sk_bound_dev_if = usin->sin6_scope_id;
880 }
881
882 /* Connect to link-local address requires an interface */
883 if (!sk->sk_bound_dev_if)
884 return -EINVAL;
885 }
886
887 sk->sk_v6_daddr = usin->sin6_addr;
888 np->flow_label = fl6.flowlabel;
889
890 /*
891 * DCCP over IPv4
892 */
893 if (addr_type == IPV6_ADDR_MAPPED) {
894 u32 exthdrlen = icsk->icsk_ext_hdr_len;
895 struct sockaddr_in sin;
896
897 net_dbg_ratelimited("connect: ipv4 mapped\n");
898
899 if (ipv6_only_sock(sk))
900 return -ENETUNREACH;
901
902 sin.sin_family = AF_INET;
903 sin.sin_port = usin->sin6_port;
904 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
905
906 icsk->icsk_af_ops = &dccp_ipv6_mapped;
907 sk->sk_backlog_rcv = dccp_v4_do_rcv;
908
909 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
910 if (err) {
911 icsk->icsk_ext_hdr_len = exthdrlen;
912 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
913 sk->sk_backlog_rcv = dccp_v6_do_rcv;
914 goto failure;
915 }
916 np->saddr = sk->sk_v6_rcv_saddr;
917 return err;
918 }
919
920 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
921 saddr = &sk->sk_v6_rcv_saddr;
922
923 fl6.flowi6_proto = IPPROTO_DCCP;
924 fl6.daddr = sk->sk_v6_daddr;
925 fl6.saddr = saddr ? *saddr : np->saddr;
926 fl6.flowi6_oif = sk->sk_bound_dev_if;
927 fl6.fl6_dport = usin->sin6_port;
928 fl6.fl6_sport = inet->inet_sport;
929 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
930
931 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
932 final_p = fl6_update_dst(&fl6, opt, &final);
933
934 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
935 if (IS_ERR(dst)) {
936 err = PTR_ERR(dst);
937 goto failure;
938 }
939
940 if (saddr == NULL) {
941 saddr = &fl6.saddr;
942
943 err = inet_bhash2_update_saddr(sk, saddr, AF_INET6);
944 if (err)
945 goto failure;
946 }
947
948 /* set the source address */
949 np->saddr = *saddr;
950 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
951
952 ip6_dst_store(sk, dst, NULL, NULL);
953
954 icsk->icsk_ext_hdr_len = 0;
955 if (opt)
956 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
957
958 inet->inet_dport = usin->sin6_port;
959
960 dccp_set_state(sk, DCCP_REQUESTING);
961 err = inet6_hash_connect(&dccp_death_row, sk);
962 if (err)
963 goto late_failure;
964
965 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
966 sk->sk_v6_daddr.s6_addr32,
967 inet->inet_sport,
968 inet->inet_dport);
969 err = dccp_connect(sk);
970 if (err)
971 goto late_failure;
972
973 return 0;
974
975 late_failure:
976 dccp_set_state(sk, DCCP_CLOSED);
977 inet_bhash2_reset_saddr(sk);
978 __sk_dst_reset(sk);
979 failure:
980 inet->inet_dport = 0;
981 sk->sk_route_caps = 0;
982 return err;
983 }
984
985 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
986 .queue_xmit = inet6_csk_xmit,
987 .send_check = dccp_v6_send_check,
988 .rebuild_header = inet6_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 ipv6hdr),
992 .setsockopt = ipv6_setsockopt,
993 .getsockopt = ipv6_getsockopt,
994 .addr2sockaddr = inet6_csk_addr2sockaddr,
995 .sockaddr_len = sizeof(struct sockaddr_in6),
996 };
997
998 /*
999 * DCCP over IPv4 via INET6 API
1000 */
1001 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1002 .queue_xmit = ip_queue_xmit,
1003 .send_check = dccp_v4_send_check,
1004 .rebuild_header = inet_sk_rebuild_header,
1005 .conn_request = dccp_v6_conn_request,
1006 .syn_recv_sock = dccp_v6_request_recv_sock,
1007 .net_header_len = sizeof(struct iphdr),
1008 .setsockopt = ipv6_setsockopt,
1009 .getsockopt = ipv6_getsockopt,
1010 .addr2sockaddr = inet6_csk_addr2sockaddr,
1011 .sockaddr_len = sizeof(struct sockaddr_in6),
1012 };
1013
dccp_v6_sk_destruct(struct sock * sk)1014 static void dccp_v6_sk_destruct(struct sock *sk)
1015 {
1016 dccp_destruct_common(sk);
1017 inet6_sock_destruct(sk);
1018 }
1019
1020 /* NOTE: A lot of things set to zero explicitly by call to
1021 * sk_alloc() so need not be done here.
1022 */
dccp_v6_init_sock(struct sock * sk)1023 static int dccp_v6_init_sock(struct sock *sk)
1024 {
1025 static __u8 dccp_v6_ctl_sock_initialized;
1026 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1027
1028 if (err == 0) {
1029 if (unlikely(!dccp_v6_ctl_sock_initialized))
1030 dccp_v6_ctl_sock_initialized = 1;
1031 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1032 sk->sk_destruct = dccp_v6_sk_destruct;
1033 }
1034
1035 return err;
1036 }
1037
1038 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1039 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1040 };
1041
1042 static struct proto dccp_v6_prot = {
1043 .name = "DCCPv6",
1044 .owner = THIS_MODULE,
1045 .close = dccp_close,
1046 .connect = dccp_v6_connect,
1047 .disconnect = dccp_disconnect,
1048 .ioctl = dccp_ioctl,
1049 .init = dccp_v6_init_sock,
1050 .setsockopt = dccp_setsockopt,
1051 .getsockopt = dccp_getsockopt,
1052 .sendmsg = dccp_sendmsg,
1053 .recvmsg = dccp_recvmsg,
1054 .backlog_rcv = dccp_v6_do_rcv,
1055 .hash = inet6_hash,
1056 .unhash = inet_unhash,
1057 .accept = inet_csk_accept,
1058 .get_port = inet_csk_get_port,
1059 .shutdown = dccp_shutdown,
1060 .destroy = dccp_destroy_sock,
1061 .orphan_count = &dccp_orphan_count,
1062 .max_header = MAX_DCCP_HEADER,
1063 .obj_size = sizeof(struct dccp6_sock),
1064 .ipv6_pinfo_offset = offsetof(struct dccp6_sock, inet6),
1065 .slab_flags = SLAB_TYPESAFE_BY_RCU,
1066 .rsk_prot = &dccp6_request_sock_ops,
1067 .twsk_prot = &dccp6_timewait_sock_ops,
1068 .h.hashinfo = &dccp_hashinfo,
1069 };
1070
1071 static const struct inet6_protocol dccp_v6_protocol = {
1072 .handler = dccp_v6_rcv,
1073 .err_handler = dccp_v6_err,
1074 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1075 };
1076
1077 static const struct proto_ops inet6_dccp_ops = {
1078 .family = PF_INET6,
1079 .owner = THIS_MODULE,
1080 .release = inet6_release,
1081 .bind = inet6_bind,
1082 .connect = inet_stream_connect,
1083 .socketpair = sock_no_socketpair,
1084 .accept = inet_accept,
1085 .getname = inet6_getname,
1086 .poll = dccp_poll,
1087 .ioctl = inet6_ioctl,
1088 .gettstamp = sock_gettstamp,
1089 .listen = inet_dccp_listen,
1090 .shutdown = inet_shutdown,
1091 .setsockopt = sock_common_setsockopt,
1092 .getsockopt = sock_common_getsockopt,
1093 .sendmsg = inet_sendmsg,
1094 .recvmsg = sock_common_recvmsg,
1095 .mmap = sock_no_mmap,
1096 #ifdef CONFIG_COMPAT
1097 .compat_ioctl = inet6_compat_ioctl,
1098 #endif
1099 };
1100
1101 static struct inet_protosw dccp_v6_protosw = {
1102 .type = SOCK_DCCP,
1103 .protocol = IPPROTO_DCCP,
1104 .prot = &dccp_v6_prot,
1105 .ops = &inet6_dccp_ops,
1106 .flags = INET_PROTOSW_ICSK,
1107 };
1108
dccp_v6_init_net(struct net * net)1109 static int __net_init dccp_v6_init_net(struct net *net)
1110 {
1111 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1112
1113 if (dccp_hashinfo.bhash == NULL)
1114 return -ESOCKTNOSUPPORT;
1115
1116 return inet_ctl_sock_create(&pn->v6_ctl_sk, PF_INET6,
1117 SOCK_DCCP, IPPROTO_DCCP, net);
1118 }
1119
dccp_v6_exit_net(struct net * net)1120 static void __net_exit dccp_v6_exit_net(struct net *net)
1121 {
1122 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1123
1124 inet_ctl_sock_destroy(pn->v6_ctl_sk);
1125 }
1126
1127 static struct pernet_operations dccp_v6_ops = {
1128 .init = dccp_v6_init_net,
1129 .exit = dccp_v6_exit_net,
1130 .id = &dccp_v6_pernet_id,
1131 .size = sizeof(struct dccp_v6_pernet),
1132 };
1133
dccp_v6_init(void)1134 static int __init dccp_v6_init(void)
1135 {
1136 int err = proto_register(&dccp_v6_prot, 1);
1137
1138 if (err)
1139 goto out;
1140
1141 inet6_register_protosw(&dccp_v6_protosw);
1142
1143 err = register_pernet_subsys(&dccp_v6_ops);
1144 if (err)
1145 goto out_destroy_ctl_sock;
1146
1147 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1148 if (err)
1149 goto out_unregister_proto;
1150
1151 out:
1152 return err;
1153 out_unregister_proto:
1154 unregister_pernet_subsys(&dccp_v6_ops);
1155 out_destroy_ctl_sock:
1156 inet6_unregister_protosw(&dccp_v6_protosw);
1157 proto_unregister(&dccp_v6_prot);
1158 goto out;
1159 }
1160
dccp_v6_exit(void)1161 static void __exit dccp_v6_exit(void)
1162 {
1163 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1164 unregister_pernet_subsys(&dccp_v6_ops);
1165 inet6_unregister_protosw(&dccp_v6_protosw);
1166 proto_unregister(&dccp_v6_prot);
1167 }
1168
1169 module_init(dccp_v6_init);
1170 module_exit(dccp_v6_exit);
1171
1172 /*
1173 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1174 * values directly, Also cover the case where the protocol is not specified,
1175 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1176 */
1177 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1178 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1179 MODULE_LICENSE("GPL");
1180 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1181 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
1182