?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
target: refactor core_update_device_list_for_node()
[linux-2.6.git] / net / dccp / ipv6.c
blobfa9512d86f3b84f31e0a13f97d4ce9c7776fed34
1 /*
2 * DCCP over IPv6
3 * Linux INET6 implementation
4 *
5 * Based on net/dccp6/ipv6.c
6 *
7 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 */
14
15 #include <linux/module.h>
16 #include <linux/random.h>
17 #include <linux/slab.h>
18 #include <linux/xfrm.h>
19
20 #include <net/addrconf.h>
21 #include <net/inet_common.h>
22 #include <net/inet_hashtables.h>
23 #include <net/inet_sock.h>
24 #include <net/inet6_connection_sock.h>
25 #include <net/inet6_hashtables.h>
26 #include <net/ip6_route.h>
27 #include <net/ipv6.h>
28 #include <net/protocol.h>
29 #include <net/transp_v6.h>
30 #include <net/ip6_checksum.h>
31 #include <net/xfrm.h>
32 #include <net/secure_seq.h>
33
34 #include "dccp.h"
35 #include "ipv6.h"
36 #include "feat.h"
37
38 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
39
40 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
41 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
42
43 static void dccp_v6_hash(struct sock *sk)
44 {
45 if (sk->sk_state != DCCP_CLOSED) {
46 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
47 inet_hash(sk);
48 return;
49 }
50 local_bh_disable();
51 __inet6_hash(sk, NULL);
52 local_bh_enable();
53 }
54 }
55
56 /* add pseudo-header to DCCP checksum stored in skb->csum */
57 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
58 const struct in6_addr *saddr,
59 const struct in6_addr *daddr)
60 {
61 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
62 }
63
64 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
65 {
66 struct ipv6_pinfo *np = inet6_sk(sk);
67 struct dccp_hdr *dh = dccp_hdr(skb);
68
69 dccp_csum_outgoing(skb);
70 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr);
71 }
72
73 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
74 {
75 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
76 ipv6_hdr(skb)->saddr.s6_addr32,
77 dccp_hdr(skb)->dccph_dport,
78 dccp_hdr(skb)->dccph_sport );
79
80 }
81
82 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
83 u8 type, u8 code, int offset, __be32 info)
84 {
85 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
86 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
87 struct dccp_sock *dp;
88 struct ipv6_pinfo *np;
89 struct sock *sk;
90 int err;
91 __u64 seq;
92 struct net *net = dev_net(skb->dev);
93
94 if (skb->len < offset + sizeof(*dh) ||
95 skb->len < offset + __dccp_basic_hdr_len(dh)) {
96 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
97 ICMP6_MIB_INERRORS);
98 return;
99 }
100
101 sk = inet6_lookup(net, &dccp_hashinfo,
102 &hdr->daddr, dh->dccph_dport,
103 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
104
105 if (sk == NULL) {
106 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
107 ICMP6_MIB_INERRORS);
108 return;
109 }
110
111 if (sk->sk_state == DCCP_TIME_WAIT) {
112 inet_twsk_put(inet_twsk(sk));
113 return;
114 }
115
116 bh_lock_sock(sk);
117 if (sock_owned_by_user(sk))
118 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
119
120 if (sk->sk_state == DCCP_CLOSED)
121 goto out;
122
123 dp = dccp_sk(sk);
124 seq = dccp_hdr_seq(dh);
125 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
126 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
127 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
128 goto out;
129 }
130
131 np = inet6_sk(sk);
132
133 if (type == ICMPV6_PKT_TOOBIG) {
134 struct dst_entry *dst = NULL;
135
136 if (sock_owned_by_user(sk))
137 goto out;
138 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
139 goto out;
140
141 /* icmp should have updated the destination cache entry */
142 dst = __sk_dst_check(sk, np->dst_cookie);
143 if (dst == NULL) {
144 struct inet_sock *inet = inet_sk(sk);
145 struct flowi6 fl6;
146
147 /* BUGGG_FUTURE: Again, it is not clear how
148 to handle rthdr case. Ignore this complexity
149 for now.
150 */
151 memset(&fl6, 0, sizeof(fl6));
152 fl6.flowi6_proto = IPPROTO_DCCP;
153 fl6.daddr = np->daddr;
154 fl6.saddr = np->saddr;
155 fl6.flowi6_oif = sk->sk_bound_dev_if;
156 fl6.fl6_dport = inet->inet_dport;
157 fl6.fl6_sport = inet->inet_sport;
158 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
159
160 dst = ip6_dst_lookup_flow(sk, &fl6, NULL, false);
161 if (IS_ERR(dst)) {
162 sk->sk_err_soft = -PTR_ERR(dst);
163 goto out;
164 }
165 } else
166 dst_hold(dst);
167
168 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
169 dccp_sync_mss(sk, dst_mtu(dst));
170 } /* else let the usual retransmit timer handle it */
171 dst_release(dst);
172 goto out;
173 }
174
175 icmpv6_err_convert(type, code, &err);
176
177 /* Might be for an request_sock */
178 switch (sk->sk_state) {
179 struct request_sock *req, **prev;
180 case DCCP_LISTEN:
181 if (sock_owned_by_user(sk))
182 goto out;
183
184 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
185 &hdr->daddr, &hdr->saddr,
186 inet6_iif(skb));
187 if (req == NULL)
188 goto out;
189
190 /*
191 * ICMPs are not backlogged, hence we cannot get an established
192 * socket here.
193 */
194 WARN_ON(req->sk != NULL);
195
196 if (!between48(seq, dccp_rsk(req)->dreq_iss,
197 dccp_rsk(req)->dreq_gss)) {
198 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
199 goto out;
200 }
201
202 inet_csk_reqsk_queue_drop(sk, req, prev);
203 goto out;
204
205 case DCCP_REQUESTING:
206 case DCCP_RESPOND: /* Cannot happen.
207 It can, it SYNs are crossed. --ANK */
208 if (!sock_owned_by_user(sk)) {
209 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
210 sk->sk_err = err;
211 /*
212 * Wake people up to see the error
213 * (see connect in sock.c)
214 */
215 sk->sk_error_report(sk);
216 dccp_done(sk);
217 } else
218 sk->sk_err_soft = err;
219 goto out;
220 }
221
222 if (!sock_owned_by_user(sk) && np->recverr) {
223 sk->sk_err = err;
224 sk->sk_error_report(sk);
225 } else
226 sk->sk_err_soft = err;
227
228 out:
229 bh_unlock_sock(sk);
230 sock_put(sk);
231 }
232
233
234 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
235 struct request_values *rv_unused)
236 {
237 struct inet6_request_sock *ireq6 = inet6_rsk(req);
238 struct ipv6_pinfo *np = inet6_sk(sk);
239 struct sk_buff *skb;
240 struct ipv6_txoptions *opt = NULL;
241 struct in6_addr *final_p, final;
242 struct flowi6 fl6;
243 int err = -1;
244 struct dst_entry *dst;
245
246 memset(&fl6, 0, sizeof(fl6));
247 fl6.flowi6_proto = IPPROTO_DCCP;
248 fl6.daddr = ireq6->rmt_addr;
249 fl6.saddr = ireq6->loc_addr;
250 fl6.flowlabel = 0;
251 fl6.flowi6_oif = ireq6->iif;
252 fl6.fl6_dport = inet_rsk(req)->rmt_port;
253 fl6.fl6_sport = inet_rsk(req)->loc_port;
254 security_req_classify_flow(req, flowi6_to_flowi(&fl6));
255
256 opt = np->opt;
257
258 final_p = fl6_update_dst(&fl6, opt, &final);
259
260 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
261 if (IS_ERR(dst)) {
262 err = PTR_ERR(dst);
263 dst = NULL;
264 goto done;
265 }
266
267 skb = dccp_make_response(sk, dst, req);
268 if (skb != NULL) {
269 struct dccp_hdr *dh = dccp_hdr(skb);
270
271 dh->dccph_checksum = dccp_v6_csum_finish(skb,
272 &ireq6->loc_addr,
273 &ireq6->rmt_addr);
274 fl6.daddr = ireq6->rmt_addr;
275 err = ip6_xmit(sk, skb, &fl6, opt, np->tclass);
276 err = net_xmit_eval(err);
277 }
278
279 done:
280 if (opt != NULL && opt != np->opt)
281 sock_kfree_s(sk, opt, opt->tot_len);
282 dst_release(dst);
283 return err;
284 }
285
286 static void dccp_v6_reqsk_destructor(struct request_sock *req)
287 {
288 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
289 if (inet6_rsk(req)->pktopts != NULL)
290 kfree_skb(inet6_rsk(req)->pktopts);
291 }
292
293 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
294 {
295 const struct ipv6hdr *rxip6h;
296 struct sk_buff *skb;
297 struct flowi6 fl6;
298 struct net *net = dev_net(skb_dst(rxskb)->dev);
299 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
300 struct dst_entry *dst;
301
302 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
303 return;
304
305 if (!ipv6_unicast_destination(rxskb))
306 return;
307
308 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
309 if (skb == NULL)
310 return;
311
312 rxip6h = ipv6_hdr(rxskb);
313 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
314 &rxip6h->daddr);
315
316 memset(&fl6, 0, sizeof(fl6));
317 fl6.daddr = rxip6h->saddr;
318 fl6.saddr = rxip6h->daddr;
319
320 fl6.flowi6_proto = IPPROTO_DCCP;
321 fl6.flowi6_oif = inet6_iif(rxskb);
322 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
323 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
324 security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
325
326 /* sk = NULL, but it is safe for now. RST socket required. */
327 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
328 if (!IS_ERR(dst)) {
329 skb_dst_set(skb, dst);
330 ip6_xmit(ctl_sk, skb, &fl6, NULL, 0);
331 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
332 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
333 return;
334 }
335
336 kfree_skb(skb);
337 }
338
339 static struct request_sock_ops dccp6_request_sock_ops = {
340 .family = AF_INET6,
341 .obj_size = sizeof(struct dccp6_request_sock),
342 .rtx_syn_ack = dccp_v6_send_response,
343 .send_ack = dccp_reqsk_send_ack,
344 .destructor = dccp_v6_reqsk_destructor,
345 .send_reset = dccp_v6_ctl_send_reset,
346 .syn_ack_timeout = dccp_syn_ack_timeout,
347 };
348
349 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
350 {
351 const struct dccp_hdr *dh = dccp_hdr(skb);
352 const struct ipv6hdr *iph = ipv6_hdr(skb);
353 struct sock *nsk;
354 struct request_sock **prev;
355 /* Find possible connection requests. */
356 struct request_sock *req = inet6_csk_search_req(sk, &prev,
357 dh->dccph_sport,
358 &iph->saddr,
359 &iph->daddr,
360 inet6_iif(skb));
361 if (req != NULL)
362 return dccp_check_req(sk, skb, req, prev);
363
364 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
365 &iph->saddr, dh->dccph_sport,
366 &iph->daddr, ntohs(dh->dccph_dport),
367 inet6_iif(skb));
368 if (nsk != NULL) {
369 if (nsk->sk_state != DCCP_TIME_WAIT) {
370 bh_lock_sock(nsk);
371 return nsk;
372 }
373 inet_twsk_put(inet_twsk(nsk));
374 return NULL;
375 }
376
377 return sk;
378 }
379
380 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
381 {
382 struct request_sock *req;
383 struct dccp_request_sock *dreq;
384 struct inet6_request_sock *ireq6;
385 struct ipv6_pinfo *np = inet6_sk(sk);
386 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
387 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
388
389 if (skb->protocol == htons(ETH_P_IP))
390 return dccp_v4_conn_request(sk, skb);
391
392 if (!ipv6_unicast_destination(skb))
393 return 0; /* discard, don't send a reset here */
394
395 if (dccp_bad_service_code(sk, service)) {
396 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
397 goto drop;
398 }
399 /*
400 * There are no SYN attacks on IPv6, yet...
401 */
402 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
403 if (inet_csk_reqsk_queue_is_full(sk))
404 goto drop;
405
406 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
407 goto drop;
408
409 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
410 if (req == NULL)
411 goto drop;
412
413 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
414 goto drop_and_free;
415
416 dreq = dccp_rsk(req);
417 if (dccp_parse_options(sk, dreq, skb))
418 goto drop_and_free;
419
420 if (security_inet_conn_request(sk, skb, req))
421 goto drop_and_free;
422
423 ireq6 = inet6_rsk(req);
424 ireq6->rmt_addr = ipv6_hdr(skb)->saddr;
425 ireq6->loc_addr = ipv6_hdr(skb)->daddr;
426
427 if (ipv6_opt_accepted(sk, skb) ||
428 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
429 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
430 atomic_inc(&skb->users);
431 ireq6->pktopts = skb;
432 }
433 ireq6->iif = sk->sk_bound_dev_if;
434
435 /* So that link locals have meaning */
436 if (!sk->sk_bound_dev_if &&
437 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
438 ireq6->iif = inet6_iif(skb);
439
440 /*
441 * Step 3: Process LISTEN state
442 *
443 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
444 *
445 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
446 */
447 dreq->dreq_isr = dcb->dccpd_seq;
448 dreq->dreq_gsr = dreq->dreq_isr;
449 dreq->dreq_iss = dccp_v6_init_sequence(skb);
450 dreq->dreq_gss = dreq->dreq_iss;
451 dreq->dreq_service = service;
452
453 if (dccp_v6_send_response(sk, req, NULL))
454 goto drop_and_free;
455
456 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
457 return 0;
458
459 drop_and_free:
460 reqsk_free(req);
461 drop:
462 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
463 return -1;
464 }
465
466 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
467 struct sk_buff *skb,
468 struct request_sock *req,
469 struct dst_entry *dst)
470 {
471 struct inet6_request_sock *ireq6 = inet6_rsk(req);
472 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
473 struct inet_sock *newinet;
474 struct dccp6_sock *newdp6;
475 struct sock *newsk;
476 struct ipv6_txoptions *opt;
477
478 if (skb->protocol == htons(ETH_P_IP)) {
479 /*
480 * v6 mapped
481 */
482 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
483 if (newsk == NULL)
484 return NULL;
485
486 newdp6 = (struct dccp6_sock *)newsk;
487 newinet = inet_sk(newsk);
488 newinet->pinet6 = &newdp6->inet6;
489 newnp = inet6_sk(newsk);
490
491 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
492
493 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
494
495 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
496
497 newnp->rcv_saddr = newnp->saddr;
498
499 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
500 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
501 newnp->pktoptions = NULL;
502 newnp->opt = NULL;
503 newnp->mcast_oif = inet6_iif(skb);
504 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
505
506 /*
507 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
508 * here, dccp_create_openreq_child now does this for us, see the comment in
509 * that function for the gory details. -acme
510 */
511
512 /* It is tricky place. Until this moment IPv4 tcp
513 worked with IPv6 icsk.icsk_af_ops.
514 Sync it now.
515 */
516 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
517
518 return newsk;
519 }
520
521 opt = np->opt;
522
523 if (sk_acceptq_is_full(sk))
524 goto out_overflow;
525
526 if (dst == NULL) {
527 struct in6_addr *final_p, final;
528 struct flowi6 fl6;
529
530 memset(&fl6, 0, sizeof(fl6));
531 fl6.flowi6_proto = IPPROTO_DCCP;
532 fl6.daddr = ireq6->rmt_addr;
533 final_p = fl6_update_dst(&fl6, opt, &final);
534 fl6.saddr = ireq6->loc_addr;
535 fl6.flowi6_oif = sk->sk_bound_dev_if;
536 fl6.fl6_dport = inet_rsk(req)->rmt_port;
537 fl6.fl6_sport = inet_rsk(req)->loc_port;
538 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
539
540 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
541 if (IS_ERR(dst))
542 goto out;
543 }
544
545 newsk = dccp_create_openreq_child(sk, req, skb);
546 if (newsk == NULL)
547 goto out_nonewsk;
548
549 /*
550 * No need to charge this sock to the relevant IPv6 refcnt debug socks
551 * count here, dccp_create_openreq_child now does this for us, see the
552 * comment in that function for the gory details. -acme
553 */
554
555 __ip6_dst_store(newsk, dst, NULL, NULL);
556 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
557 NETIF_F_TSO);
558 newdp6 = (struct dccp6_sock *)newsk;
559 newinet = inet_sk(newsk);
560 newinet->pinet6 = &newdp6->inet6;
561 newnp = inet6_sk(newsk);
562
563 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
564
565 newnp->daddr = ireq6->rmt_addr;
566 newnp->saddr = ireq6->loc_addr;
567 newnp->rcv_saddr = ireq6->loc_addr;
568 newsk->sk_bound_dev_if = ireq6->iif;
569
570 /* Now IPv6 options...
571
572 First: no IPv4 options.
573 */
574 newinet->inet_opt = NULL;
575
576 /* Clone RX bits */
577 newnp->rxopt.all = np->rxopt.all;
578
579 /* Clone pktoptions received with SYN */
580 newnp->pktoptions = NULL;
581 if (ireq6->pktopts != NULL) {
582 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
583 consume_skb(ireq6->pktopts);
584 ireq6->pktopts = NULL;
585 if (newnp->pktoptions)
586 skb_set_owner_r(newnp->pktoptions, newsk);
587 }
588 newnp->opt = NULL;
589 newnp->mcast_oif = inet6_iif(skb);
590 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
591
592 /*
593 * Clone native IPv6 options from listening socket (if any)
594 *
595 * Yes, keeping reference count would be much more clever, but we make
596 * one more one thing there: reattach optmem to newsk.
597 */
598 if (opt != NULL) {
599 newnp->opt = ipv6_dup_options(newsk, opt);
600 if (opt != np->opt)
601 sock_kfree_s(sk, opt, opt->tot_len);
602 }
603
604 inet_csk(newsk)->icsk_ext_hdr_len = 0;
605 if (newnp->opt != NULL)
606 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
607 newnp->opt->opt_flen);
608
609 dccp_sync_mss(newsk, dst_mtu(dst));
610
611 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
612 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
613
614 if (__inet_inherit_port(sk, newsk) < 0) {
615 sock_put(newsk);
616 goto out;
617 }
618 __inet6_hash(newsk, NULL);
619
620 return newsk;
621
622 out_overflow:
623 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
624 out_nonewsk:
625 dst_release(dst);
626 out:
627 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
628 if (opt != NULL && opt != np->opt)
629 sock_kfree_s(sk, opt, opt->tot_len);
630 return NULL;
631 }
632
633 /* The socket must have it's spinlock held when we get
634 * here.
635 *
636 * We have a potential double-lock case here, so even when
637 * doing backlog processing we use the BH locking scheme.
638 * This is because we cannot sleep with the original spinlock
639 * held.
640 */
641 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
642 {
643 struct ipv6_pinfo *np = inet6_sk(sk);
644 struct sk_buff *opt_skb = NULL;
645
646 /* Imagine: socket is IPv6. IPv4 packet arrives,
647 goes to IPv4 receive handler and backlogged.
648 From backlog it always goes here. Kerboom...
649 Fortunately, dccp_rcv_established and rcv_established
650 handle them correctly, but it is not case with
651 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
652 */
653
654 if (skb->protocol == htons(ETH_P_IP))
655 return dccp_v4_do_rcv(sk, skb);
656
657 if (sk_filter(sk, skb))
658 goto discard;
659
660 /*
661 * socket locking is here for SMP purposes as backlog rcv is currently
662 * called with bh processing disabled.
663 */
664
665 /* Do Stevens' IPV6_PKTOPTIONS.
666
667 Yes, guys, it is the only place in our code, where we
668 may make it not affecting IPv4.
669 The rest of code is protocol independent,
670 and I do not like idea to uglify IPv4.
671
672 Actually, all the idea behind IPV6_PKTOPTIONS
673 looks not very well thought. For now we latch
674 options, received in the last packet, enqueued
675 by tcp. Feel free to propose better solution.
676 --ANK (980728)
677 */
678 if (np->rxopt.all)
679 /*
680 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
681 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
682 */
683 opt_skb = skb_clone(skb, GFP_ATOMIC);
684
685 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
686 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
687 goto reset;
688 if (opt_skb) {
689 /* XXX This is where we would goto ipv6_pktoptions. */
690 __kfree_skb(opt_skb);
691 }
692 return 0;
693 }
694
695 /*
696 * Step 3: Process LISTEN state
697 * If S.state == LISTEN,
698 * If P.type == Request or P contains a valid Init Cookie option,
699 * (* Must scan the packet's options to check for Init
700 * Cookies. Only Init Cookies are processed here,
701 * however; other options are processed in Step 8. This
702 * scan need only be performed if the endpoint uses Init
703 * Cookies *)
704 * (* Generate a new socket and switch to that socket *)
705 * Set S := new socket for this port pair
706 * S.state = RESPOND
707 * Choose S.ISS (initial seqno) or set from Init Cookies
708 * Initialize S.GAR := S.ISS
709 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
710 * Continue with S.state == RESPOND
711 * (* A Response packet will be generated in Step 11 *)
712 * Otherwise,
713 * Generate Reset(No Connection) unless P.type == Reset
714 * Drop packet and return
715 *
716 * NOTE: the check for the packet types is done in
717 * dccp_rcv_state_process
718 */
719 if (sk->sk_state == DCCP_LISTEN) {
720 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
721
722 if (nsk == NULL)
723 goto discard;
724 /*
725 * Queue it on the new socket if the new socket is active,
726 * otherwise we just shortcircuit this and continue with
727 * the new socket..
728 */
729 if (nsk != sk) {
730 if (dccp_child_process(sk, nsk, skb))
731 goto reset;
732 if (opt_skb != NULL)
733 __kfree_skb(opt_skb);
734 return 0;
735 }
736 }
737
738 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
739 goto reset;
740 if (opt_skb) {
741 /* XXX This is where we would goto ipv6_pktoptions. */
742 __kfree_skb(opt_skb);
743 }
744 return 0;
745
746 reset:
747 dccp_v6_ctl_send_reset(sk, skb);
748 discard:
749 if (opt_skb != NULL)
750 __kfree_skb(opt_skb);
751 kfree_skb(skb);
752 return 0;
753 }
754
755 static int dccp_v6_rcv(struct sk_buff *skb)
756 {
757 const struct dccp_hdr *dh;
758 struct sock *sk;
759 int min_cov;
760
761 /* Step 1: Check header basics */
762
763 if (dccp_invalid_packet(skb))
764 goto discard_it;
765
766 /* Step 1: If header checksum is incorrect, drop packet and return. */
767 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
768 &ipv6_hdr(skb)->daddr)) {
769 DCCP_WARN("dropped packet with invalid checksum\n");
770 goto discard_it;
771 }
772
773 dh = dccp_hdr(skb);
774
775 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
776 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
777
778 if (dccp_packet_without_ack(skb))
779 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
780 else
781 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
782
783 /* Step 2:
784 * Look up flow ID in table and get corresponding socket */
785 sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
786 dh->dccph_sport, dh->dccph_dport);
787 /*
788 * Step 2:
789 * If no socket ...
790 */
791 if (sk == NULL) {
792 dccp_pr_debug("failed to look up flow ID in table and "
793 "get corresponding socket\n");
794 goto no_dccp_socket;
795 }
796
797 /*
798 * Step 2:
799 * ... or S.state == TIMEWAIT,
800 * Generate Reset(No Connection) unless P.type == Reset
801 * Drop packet and return
802 */
803 if (sk->sk_state == DCCP_TIME_WAIT) {
804 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
805 inet_twsk_put(inet_twsk(sk));
806 goto no_dccp_socket;
807 }
808
809 /*
810 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
811 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
812 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
813 */
814 min_cov = dccp_sk(sk)->dccps_pcrlen;
815 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
816 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
817 dh->dccph_cscov, min_cov);
818 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
819 goto discard_and_relse;
820 }
821
822 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
823 goto discard_and_relse;
824
825 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
826
827 no_dccp_socket:
828 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
829 goto discard_it;
830 /*
831 * Step 2:
832 * If no socket ...
833 * Generate Reset(No Connection) unless P.type == Reset
834 * Drop packet and return
835 */
836 if (dh->dccph_type != DCCP_PKT_RESET) {
837 DCCP_SKB_CB(skb)->dccpd_reset_code =
838 DCCP_RESET_CODE_NO_CONNECTION;
839 dccp_v6_ctl_send_reset(sk, skb);
840 }
841
842 discard_it:
843 kfree_skb(skb);
844 return 0;
845
846 discard_and_relse:
847 sock_put(sk);
848 goto discard_it;
849 }
850
851 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
852 int addr_len)
853 {
854 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
855 struct inet_connection_sock *icsk = inet_csk(sk);
856 struct inet_sock *inet = inet_sk(sk);
857 struct ipv6_pinfo *np = inet6_sk(sk);
858 struct dccp_sock *dp = dccp_sk(sk);
859 struct in6_addr *saddr = NULL, *final_p, final;
860 struct flowi6 fl6;
861 struct dst_entry *dst;
862 int addr_type;
863 int err;
864
865 dp->dccps_role = DCCP_ROLE_CLIENT;
866
867 if (addr_len < SIN6_LEN_RFC2133)
868 return -EINVAL;
869
870 if (usin->sin6_family != AF_INET6)
871 return -EAFNOSUPPORT;
872
873 memset(&fl6, 0, sizeof(fl6));
874
875 if (np->sndflow) {
876 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
877 IP6_ECN_flow_init(fl6.flowlabel);
878 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
879 struct ip6_flowlabel *flowlabel;
880 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
881 if (flowlabel == NULL)
882 return -EINVAL;
883 usin->sin6_addr = flowlabel->dst;
884 fl6_sock_release(flowlabel);
885 }
886 }
887 /*
888 * connect() to INADDR_ANY means loopback (BSD'ism).
889 */
890 if (ipv6_addr_any(&usin->sin6_addr))
891 usin->sin6_addr.s6_addr[15] = 1;
892
893 addr_type = ipv6_addr_type(&usin->sin6_addr);
894
895 if (addr_type & IPV6_ADDR_MULTICAST)
896 return -ENETUNREACH;
897
898 if (addr_type & IPV6_ADDR_LINKLOCAL) {
899 if (addr_len >= sizeof(struct sockaddr_in6) &&
900 usin->sin6_scope_id) {
901 /* If interface is set while binding, indices
902 * must coincide.
903 */
904 if (sk->sk_bound_dev_if &&
905 sk->sk_bound_dev_if != usin->sin6_scope_id)
906 return -EINVAL;
907
908 sk->sk_bound_dev_if = usin->sin6_scope_id;
909 }
910
911 /* Connect to link-local address requires an interface */
912 if (!sk->sk_bound_dev_if)
913 return -EINVAL;
914 }
915
916 np->daddr = usin->sin6_addr;
917 np->flow_label = fl6.flowlabel;
918
919 /*
920 * DCCP over IPv4
921 */
922 if (addr_type == IPV6_ADDR_MAPPED) {
923 u32 exthdrlen = icsk->icsk_ext_hdr_len;
924 struct sockaddr_in sin;
925
926 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
927
928 if (__ipv6_only_sock(sk))
929 return -ENETUNREACH;
930
931 sin.sin_family = AF_INET;
932 sin.sin_port = usin->sin6_port;
933 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
934
935 icsk->icsk_af_ops = &dccp_ipv6_mapped;
936 sk->sk_backlog_rcv = dccp_v4_do_rcv;
937
938 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
939 if (err) {
940 icsk->icsk_ext_hdr_len = exthdrlen;
941 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
942 sk->sk_backlog_rcv = dccp_v6_do_rcv;
943 goto failure;
944 }
945 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
946 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr);
947
948 return err;
949 }
950
951 if (!ipv6_addr_any(&np->rcv_saddr))
952 saddr = &np->rcv_saddr;
953
954 fl6.flowi6_proto = IPPROTO_DCCP;
955 fl6.daddr = np->daddr;
956 fl6.saddr = saddr ? *saddr : np->saddr;
957 fl6.flowi6_oif = sk->sk_bound_dev_if;
958 fl6.fl6_dport = usin->sin6_port;
959 fl6.fl6_sport = inet->inet_sport;
960 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
961
962 final_p = fl6_update_dst(&fl6, np->opt, &final);
963
964 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
965 if (IS_ERR(dst)) {
966 err = PTR_ERR(dst);
967 goto failure;
968 }
969
970 if (saddr == NULL) {
971 saddr = &fl6.saddr;
972 np->rcv_saddr = *saddr;
973 }
974
975 /* set the source address */
976 np->saddr = *saddr;
977 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
978
979 __ip6_dst_store(sk, dst, NULL, NULL);
980
981 icsk->icsk_ext_hdr_len = 0;
982 if (np->opt != NULL)
983 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
984 np->opt->opt_nflen);
985
986 inet->inet_dport = usin->sin6_port;
987
988 dccp_set_state(sk, DCCP_REQUESTING);
989 err = inet6_hash_connect(&dccp_death_row, sk);
990 if (err)
991 goto late_failure;
992
993 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
994 np->daddr.s6_addr32,
995 inet->inet_sport,
996 inet->inet_dport);
997 err = dccp_connect(sk);
998 if (err)
999 goto late_failure;
1000
1001 return 0;
1002
1003 late_failure:
1004 dccp_set_state(sk, DCCP_CLOSED);
1005 __sk_dst_reset(sk);
1006 failure:
1007 inet->inet_dport = 0;
1008 sk->sk_route_caps = 0;
1009 return err;
1010 }
1011
1012 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1013 .queue_xmit = inet6_csk_xmit,
1014 .send_check = dccp_v6_send_check,
1015 .rebuild_header = inet6_sk_rebuild_header,
1016 .conn_request = dccp_v6_conn_request,
1017 .syn_recv_sock = dccp_v6_request_recv_sock,
1018 .net_header_len = sizeof(struct ipv6hdr),
1019 .setsockopt = ipv6_setsockopt,
1020 .getsockopt = ipv6_getsockopt,
1021 .addr2sockaddr = inet6_csk_addr2sockaddr,
1022 .sockaddr_len = sizeof(struct sockaddr_in6),
1023 .bind_conflict = inet6_csk_bind_conflict,
1024 #ifdef CONFIG_COMPAT
1025 .compat_setsockopt = compat_ipv6_setsockopt,
1026 .compat_getsockopt = compat_ipv6_getsockopt,
1027 #endif
1028 };
1029
1030 /*
1031 * DCCP over IPv4 via INET6 API
1032 */
1033 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1034 .queue_xmit = ip_queue_xmit,
1035 .send_check = dccp_v4_send_check,
1036 .rebuild_header = inet_sk_rebuild_header,
1037 .conn_request = dccp_v6_conn_request,
1038 .syn_recv_sock = dccp_v6_request_recv_sock,
1039 .net_header_len = sizeof(struct iphdr),
1040 .setsockopt = ipv6_setsockopt,
1041 .getsockopt = ipv6_getsockopt,
1042 .addr2sockaddr = inet6_csk_addr2sockaddr,
1043 .sockaddr_len = sizeof(struct sockaddr_in6),
1044 #ifdef CONFIG_COMPAT
1045 .compat_setsockopt = compat_ipv6_setsockopt,
1046 .compat_getsockopt = compat_ipv6_getsockopt,
1047 #endif
1048 };
1049
1050 /* NOTE: A lot of things set to zero explicitly by call to
1051 * sk_alloc() so need not be done here.
1052 */
1053 static int dccp_v6_init_sock(struct sock *sk)
1054 {
1055 static __u8 dccp_v6_ctl_sock_initialized;
1056 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1057
1058 if (err == 0) {
1059 if (unlikely(!dccp_v6_ctl_sock_initialized))
1060 dccp_v6_ctl_sock_initialized = 1;
1061 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1062 }
1063
1064 return err;
1065 }
1066
1067 static void dccp_v6_destroy_sock(struct sock *sk)
1068 {
1069 dccp_destroy_sock(sk);
1070 inet6_destroy_sock(sk);
1071 }
1072
1073 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1074 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1075 };
1076
1077 static struct proto dccp_v6_prot = {
1078 .name = "DCCPv6",
1079 .owner = THIS_MODULE,
1080 .close = dccp_close,
1081 .connect = dccp_v6_connect,
1082 .disconnect = dccp_disconnect,
1083 .ioctl = dccp_ioctl,
1084 .init = dccp_v6_init_sock,
1085 .setsockopt = dccp_setsockopt,
1086 .getsockopt = dccp_getsockopt,
1087 .sendmsg = dccp_sendmsg,
1088 .recvmsg = dccp_recvmsg,
1089 .backlog_rcv = dccp_v6_do_rcv,
1090 .hash = dccp_v6_hash,
1091 .unhash = inet_unhash,
1092 .accept = inet_csk_accept,
1093 .get_port = inet_csk_get_port,
1094 .shutdown = dccp_shutdown,
1095 .destroy = dccp_v6_destroy_sock,
1096 .orphan_count = &dccp_orphan_count,
1097 .max_header = MAX_DCCP_HEADER,
1098 .obj_size = sizeof(struct dccp6_sock),
1099 .slab_flags = SLAB_DESTROY_BY_RCU,
1100 .rsk_prot = &dccp6_request_sock_ops,
1101 .twsk_prot = &dccp6_timewait_sock_ops,
1102 .h.hashinfo = &dccp_hashinfo,
1103 #ifdef CONFIG_COMPAT
1104 .compat_setsockopt = compat_dccp_setsockopt,
1105 .compat_getsockopt = compat_dccp_getsockopt,
1106 #endif
1107 };
1108
1109 static const struct inet6_protocol dccp_v6_protocol = {
1110 .handler = dccp_v6_rcv,
1111 .err_handler = dccp_v6_err,
1112 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1113 };
1114
1115 static const struct proto_ops inet6_dccp_ops = {
1116 .family = PF_INET6,
1117 .owner = THIS_MODULE,
1118 .release = inet6_release,
1119 .bind = inet6_bind,
1120 .connect = inet_stream_connect,
1121 .socketpair = sock_no_socketpair,
1122 .accept = inet_accept,
1123 .getname = inet6_getname,
1124 .poll = dccp_poll,
1125 .ioctl = inet6_ioctl,
1126 .listen = inet_dccp_listen,
1127 .shutdown = inet_shutdown,
1128 .setsockopt = sock_common_setsockopt,
1129 .getsockopt = sock_common_getsockopt,
1130 .sendmsg = inet_sendmsg,
1131 .recvmsg = sock_common_recvmsg,
1132 .mmap = sock_no_mmap,
1133 .sendpage = sock_no_sendpage,
1134 #ifdef CONFIG_COMPAT
1135 .compat_setsockopt = compat_sock_common_setsockopt,
1136 .compat_getsockopt = compat_sock_common_getsockopt,
1137 #endif
1138 };
1139
1140 static struct inet_protosw dccp_v6_protosw = {
1141 .type = SOCK_DCCP,
1142 .protocol = IPPROTO_DCCP,
1143 .prot = &dccp_v6_prot,
1144 .ops = &inet6_dccp_ops,
1145 .flags = INET_PROTOSW_ICSK,
1146 };
1147
1148 static int __net_init dccp_v6_init_net(struct net *net)
1149 {
1150 if (dccp_hashinfo.bhash == NULL)
1151 return -ESOCKTNOSUPPORT;
1152
1153 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1154 SOCK_DCCP, IPPROTO_DCCP, net);
1155 }
1156
1157 static void __net_exit dccp_v6_exit_net(struct net *net)
1158 {
1159 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1160 }
1161
1162 static struct pernet_operations dccp_v6_ops = {
1163 .init = dccp_v6_init_net,
1164 .exit = dccp_v6_exit_net,
1165 };
1166
1167 static int __init dccp_v6_init(void)
1168 {
1169 int err = proto_register(&dccp_v6_prot, 1);
1170
1171 if (err != 0)
1172 goto out;
1173
1174 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1175 if (err != 0)
1176 goto out_unregister_proto;
1177
1178 inet6_register_protosw(&dccp_v6_protosw);
1179
1180 err = register_pernet_subsys(&dccp_v6_ops);
1181 if (err != 0)
1182 goto out_destroy_ctl_sock;
1183 out:
1184 return err;
1185
1186 out_destroy_ctl_sock:
1187 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1188 inet6_unregister_protosw(&dccp_v6_protosw);
1189 out_unregister_proto:
1190 proto_unregister(&dccp_v6_prot);
1191 goto out;
1192 }
1193
1194 static void __exit dccp_v6_exit(void)
1195 {
1196 unregister_pernet_subsys(&dccp_v6_ops);
1197 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1198 inet6_unregister_protosw(&dccp_v6_protosw);
1199 proto_unregister(&dccp_v6_prot);
1200 }
1201
1202 module_init(dccp_v6_init);
1203 module_exit(dccp_v6_exit);
1204
1205 /*
1206 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1207 * values directly, Also cover the case where the protocol is not specified,
1208 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1209 */
1210 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1211 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1212 MODULE_LICENSE("GPL");
1213 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1214 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
Linus' kernel tree