init from v2.6.32.60
[mach-moxart.git] / net / dccp / ipv6.c
blob9ed19628da48c798512782c97e7c6027f93c2070
1 /*
2 * DCCP over IPv6
3 * Linux INET6 implementation
5 * Based on net/dccp6/ipv6.c
7 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
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.
15 #include <linux/module.h>
16 #include <linux/random.h>
17 #include <linux/xfrm.h>
19 #include <net/addrconf.h>
20 #include <net/inet_common.h>
21 #include <net/inet_hashtables.h>
22 #include <net/inet_sock.h>
23 #include <net/inet6_connection_sock.h>
24 #include <net/inet6_hashtables.h>
25 #include <net/ip6_route.h>
26 #include <net/ipv6.h>
27 #include <net/protocol.h>
28 #include <net/transp_v6.h>
29 #include <net/ip6_checksum.h>
30 #include <net/xfrm.h>
31 #include <net/secure_seq.h>
33 #include "dccp.h"
34 #include "ipv6.h"
35 #include "feat.h"
37 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
39 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
40 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
42 static void dccp_v6_hash(struct sock *sk)
44 if (sk->sk_state != DCCP_CLOSED) {
45 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
46 inet_hash(sk);
47 return;
49 local_bh_disable();
50 __inet6_hash(sk);
51 local_bh_enable();
55 /* add pseudo-header to DCCP checksum stored in skb->csum */
56 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
57 struct in6_addr *saddr,
58 struct in6_addr *daddr)
60 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
63 static inline void dccp_v6_send_check(struct sock *sk, int unused_value,
64 struct sk_buff *skb)
66 struct ipv6_pinfo *np = inet6_sk(sk);
67 struct dccp_hdr *dh = dccp_hdr(skb);
69 dccp_csum_outgoing(skb);
70 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr);
73 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
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 );
82 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
83 u8 type, u8 code, int offset, __be32 info)
85 struct ipv6hdr *hdr = (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);
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;
101 sk = inet6_lookup(net, &dccp_hashinfo,
102 &hdr->daddr, dh->dccph_dport,
103 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
105 if (sk == NULL) {
106 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
107 ICMP6_MIB_INERRORS);
108 return;
111 if (sk->sk_state == DCCP_TIME_WAIT) {
112 inet_twsk_put(inet_twsk(sk));
113 return;
116 bh_lock_sock(sk);
117 if (sock_owned_by_user(sk))
118 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
120 if (sk->sk_state == DCCP_CLOSED)
121 goto out;
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;
131 np = inet6_sk(sk);
133 if (type == ICMPV6_PKT_TOOBIG) {
134 struct dst_entry *dst = NULL;
136 if (sock_owned_by_user(sk))
137 goto out;
138 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
139 goto out;
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 flowi fl;
147 /* BUGGG_FUTURE: Again, it is not clear how
148 to handle rthdr case. Ignore this complexity
149 for now.
151 memset(&fl, 0, sizeof(fl));
152 fl.proto = IPPROTO_DCCP;
153 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
154 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
155 fl.oif = sk->sk_bound_dev_if;
156 fl.fl_ip_dport = inet->dport;
157 fl.fl_ip_sport = inet->sport;
158 security_sk_classify_flow(sk, &fl);
160 err = ip6_dst_lookup(sk, &dst, &fl);
161 if (err) {
162 sk->sk_err_soft = -err;
163 goto out;
166 err = xfrm_lookup(net, &dst, &fl, sk, 0);
167 if (err < 0) {
168 sk->sk_err_soft = -err;
169 goto out;
171 } else
172 dst_hold(dst);
174 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
175 dccp_sync_mss(sk, dst_mtu(dst));
176 } /* else let the usual retransmit timer handle it */
177 dst_release(dst);
178 goto out;
181 icmpv6_err_convert(type, code, &err);
183 /* Might be for an request_sock */
184 switch (sk->sk_state) {
185 struct request_sock *req, **prev;
186 case DCCP_LISTEN:
187 if (sock_owned_by_user(sk))
188 goto out;
190 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
191 &hdr->daddr, &hdr->saddr,
192 inet6_iif(skb));
193 if (req == NULL)
194 goto out;
197 * ICMPs are not backlogged, hence we cannot get an established
198 * socket here.
200 WARN_ON(req->sk != NULL);
202 if (seq != dccp_rsk(req)->dreq_iss) {
203 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
204 goto out;
207 inet_csk_reqsk_queue_drop(sk, req, prev);
208 goto out;
210 case DCCP_REQUESTING:
211 case DCCP_RESPOND: /* Cannot happen.
212 It can, it SYNs are crossed. --ANK */
213 if (!sock_owned_by_user(sk)) {
214 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
215 sk->sk_err = err;
217 * Wake people up to see the error
218 * (see connect in sock.c)
220 sk->sk_error_report(sk);
221 dccp_done(sk);
222 } else
223 sk->sk_err_soft = err;
224 goto out;
227 if (!sock_owned_by_user(sk) && np->recverr) {
228 sk->sk_err = err;
229 sk->sk_error_report(sk);
230 } else
231 sk->sk_err_soft = err;
233 out:
234 bh_unlock_sock(sk);
235 sock_put(sk);
239 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req)
241 struct inet6_request_sock *ireq6 = inet6_rsk(req);
242 struct ipv6_pinfo *np = inet6_sk(sk);
243 struct sk_buff *skb;
244 struct ipv6_txoptions *opt = NULL;
245 struct in6_addr *final_p = NULL, final;
246 struct flowi fl;
247 int err = -1;
248 struct dst_entry *dst;
250 memset(&fl, 0, sizeof(fl));
251 fl.proto = IPPROTO_DCCP;
252 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
253 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
254 fl.fl6_flowlabel = 0;
255 fl.oif = ireq6->iif;
256 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
257 fl.fl_ip_sport = inet_rsk(req)->loc_port;
258 security_req_classify_flow(req, &fl);
260 opt = np->opt;
262 if (opt != NULL && opt->srcrt != NULL) {
263 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
265 ipv6_addr_copy(&final, &fl.fl6_dst);
266 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
267 final_p = &final;
270 err = ip6_dst_lookup(sk, &dst, &fl);
271 if (err)
272 goto done;
274 if (final_p)
275 ipv6_addr_copy(&fl.fl6_dst, final_p);
277 err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0);
278 if (err < 0)
279 goto done;
281 skb = dccp_make_response(sk, dst, req);
282 if (skb != NULL) {
283 struct dccp_hdr *dh = dccp_hdr(skb);
285 dh->dccph_checksum = dccp_v6_csum_finish(skb,
286 &ireq6->loc_addr,
287 &ireq6->rmt_addr);
288 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
289 err = ip6_xmit(sk, skb, &fl, opt, 0);
290 err = net_xmit_eval(err);
293 done:
294 if (opt != NULL && opt != np->opt)
295 sock_kfree_s(sk, opt, opt->tot_len);
296 dst_release(dst);
297 return err;
300 static void dccp_v6_reqsk_destructor(struct request_sock *req)
302 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
303 if (inet6_rsk(req)->pktopts != NULL)
304 kfree_skb(inet6_rsk(req)->pktopts);
307 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
309 struct ipv6hdr *rxip6h;
310 struct sk_buff *skb;
311 struct flowi fl;
312 struct net *net = dev_net(skb_dst(rxskb)->dev);
313 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
314 struct dst_entry *dst;
316 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
317 return;
319 if (!ipv6_unicast_destination(rxskb))
320 return;
322 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
323 if (skb == NULL)
324 return;
326 rxip6h = ipv6_hdr(rxskb);
327 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
328 &rxip6h->daddr);
330 memset(&fl, 0, sizeof(fl));
331 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
332 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
334 fl.proto = IPPROTO_DCCP;
335 fl.oif = inet6_iif(rxskb);
336 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
337 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
338 security_skb_classify_flow(rxskb, &fl);
340 /* sk = NULL, but it is safe for now. RST socket required. */
341 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) {
342 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) {
343 skb_dst_set(skb, dst);
344 ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
345 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
346 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
347 return;
351 kfree_skb(skb);
354 static struct request_sock_ops dccp6_request_sock_ops = {
355 .family = AF_INET6,
356 .obj_size = sizeof(struct dccp6_request_sock),
357 .rtx_syn_ack = dccp_v6_send_response,
358 .send_ack = dccp_reqsk_send_ack,
359 .destructor = dccp_v6_reqsk_destructor,
360 .send_reset = dccp_v6_ctl_send_reset,
363 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
365 const struct dccp_hdr *dh = dccp_hdr(skb);
366 const struct ipv6hdr *iph = ipv6_hdr(skb);
367 struct sock *nsk;
368 struct request_sock **prev;
369 /* Find possible connection requests. */
370 struct request_sock *req = inet6_csk_search_req(sk, &prev,
371 dh->dccph_sport,
372 &iph->saddr,
373 &iph->daddr,
374 inet6_iif(skb));
375 if (req != NULL)
376 return dccp_check_req(sk, skb, req, prev);
378 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
379 &iph->saddr, dh->dccph_sport,
380 &iph->daddr, ntohs(dh->dccph_dport),
381 inet6_iif(skb));
382 if (nsk != NULL) {
383 if (nsk->sk_state != DCCP_TIME_WAIT) {
384 bh_lock_sock(nsk);
385 return nsk;
387 inet_twsk_put(inet_twsk(nsk));
388 return NULL;
391 return sk;
394 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
396 struct request_sock *req;
397 struct dccp_request_sock *dreq;
398 struct inet6_request_sock *ireq6;
399 struct ipv6_pinfo *np = inet6_sk(sk);
400 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
401 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
403 if (skb->protocol == htons(ETH_P_IP))
404 return dccp_v4_conn_request(sk, skb);
406 if (!ipv6_unicast_destination(skb))
407 return 0; /* discard, don't send a reset here */
409 if (dccp_bad_service_code(sk, service)) {
410 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
411 goto drop;
414 * There are no SYN attacks on IPv6, yet...
416 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
417 if (inet_csk_reqsk_queue_is_full(sk))
418 goto drop;
420 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
421 goto drop;
423 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
424 if (req == NULL)
425 goto drop;
427 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
428 goto drop_and_free;
430 dreq = dccp_rsk(req);
431 if (dccp_parse_options(sk, dreq, skb))
432 goto drop_and_free;
434 if (security_inet_conn_request(sk, skb, req))
435 goto drop_and_free;
437 ireq6 = inet6_rsk(req);
438 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
439 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
441 if (ipv6_opt_accepted(sk, skb) ||
442 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
443 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
444 atomic_inc(&skb->users);
445 ireq6->pktopts = skb;
447 ireq6->iif = sk->sk_bound_dev_if;
449 /* So that link locals have meaning */
450 if (!sk->sk_bound_dev_if &&
451 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
452 ireq6->iif = inet6_iif(skb);
455 * Step 3: Process LISTEN state
457 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
459 * In fact we defer setting S.GSR, S.SWL, S.SWH to
460 * dccp_create_openreq_child.
462 dreq->dreq_isr = dcb->dccpd_seq;
463 dreq->dreq_iss = dccp_v6_init_sequence(skb);
464 dreq->dreq_service = service;
466 if (dccp_v6_send_response(sk, req))
467 goto drop_and_free;
469 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
470 return 0;
472 drop_and_free:
473 reqsk_free(req);
474 drop:
475 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
476 return -1;
479 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
480 struct sk_buff *skb,
481 struct request_sock *req,
482 struct dst_entry *dst)
484 struct inet6_request_sock *ireq6 = inet6_rsk(req);
485 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
486 struct inet_sock *newinet;
487 struct dccp_sock *newdp;
488 struct dccp6_sock *newdp6;
489 struct sock *newsk;
490 struct ipv6_txoptions *opt;
492 if (skb->protocol == htons(ETH_P_IP)) {
494 * v6 mapped
496 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
497 if (newsk == NULL)
498 return NULL;
500 newdp6 = (struct dccp6_sock *)newsk;
501 newdp = dccp_sk(newsk);
502 newinet = inet_sk(newsk);
503 newinet->pinet6 = &newdp6->inet6;
504 newnp = inet6_sk(newsk);
506 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
508 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
509 newinet->daddr);
511 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
512 newinet->saddr);
514 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
516 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
517 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
518 newnp->pktoptions = NULL;
519 newnp->opt = NULL;
520 newnp->mcast_oif = inet6_iif(skb);
521 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
524 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
525 * here, dccp_create_openreq_child now does this for us, see the comment in
526 * that function for the gory details. -acme
529 /* It is tricky place. Until this moment IPv4 tcp
530 worked with IPv6 icsk.icsk_af_ops.
531 Sync it now.
533 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
535 return newsk;
538 opt = np->opt;
540 if (sk_acceptq_is_full(sk))
541 goto out_overflow;
543 if (dst == NULL) {
544 struct in6_addr *final_p = NULL, final;
545 struct flowi fl;
547 memset(&fl, 0, sizeof(fl));
548 fl.proto = IPPROTO_DCCP;
549 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
550 if (opt != NULL && opt->srcrt != NULL) {
551 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
553 ipv6_addr_copy(&final, &fl.fl6_dst);
554 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
555 final_p = &final;
557 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
558 fl.oif = sk->sk_bound_dev_if;
559 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
560 fl.fl_ip_sport = inet_rsk(req)->loc_port;
561 security_sk_classify_flow(sk, &fl);
563 if (ip6_dst_lookup(sk, &dst, &fl))
564 goto out;
566 if (final_p)
567 ipv6_addr_copy(&fl.fl6_dst, final_p);
569 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
570 goto out;
573 newsk = dccp_create_openreq_child(sk, req, skb);
574 if (newsk == NULL)
575 goto out;
578 * No need to charge this sock to the relevant IPv6 refcnt debug socks
579 * count here, dccp_create_openreq_child now does this for us, see the
580 * comment in that function for the gory details. -acme
583 __ip6_dst_store(newsk, dst, NULL, NULL);
584 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
585 NETIF_F_TSO);
586 newdp6 = (struct dccp6_sock *)newsk;
587 newinet = inet_sk(newsk);
588 newinet->pinet6 = &newdp6->inet6;
589 newdp = dccp_sk(newsk);
590 newnp = inet6_sk(newsk);
592 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
594 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
595 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
596 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
597 newsk->sk_bound_dev_if = ireq6->iif;
599 /* Now IPv6 options...
601 First: no IPv4 options.
603 newinet->opt = NULL;
605 /* Clone RX bits */
606 newnp->rxopt.all = np->rxopt.all;
608 /* Clone pktoptions received with SYN */
609 newnp->pktoptions = NULL;
610 if (ireq6->pktopts != NULL) {
611 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
612 kfree_skb(ireq6->pktopts);
613 ireq6->pktopts = NULL;
614 if (newnp->pktoptions)
615 skb_set_owner_r(newnp->pktoptions, newsk);
617 newnp->opt = NULL;
618 newnp->mcast_oif = inet6_iif(skb);
619 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
622 * Clone native IPv6 options from listening socket (if any)
624 * Yes, keeping reference count would be much more clever, but we make
625 * one more one thing there: reattach optmem to newsk.
627 if (opt != NULL) {
628 newnp->opt = ipv6_dup_options(newsk, opt);
629 if (opt != np->opt)
630 sock_kfree_s(sk, opt, opt->tot_len);
633 inet_csk(newsk)->icsk_ext_hdr_len = 0;
634 if (newnp->opt != NULL)
635 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
636 newnp->opt->opt_flen);
638 dccp_sync_mss(newsk, dst_mtu(dst));
640 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
642 __inet6_hash(newsk);
643 __inet_inherit_port(sk, newsk);
645 return newsk;
647 out_overflow:
648 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
649 out:
650 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
651 if (opt != NULL && opt != np->opt)
652 sock_kfree_s(sk, opt, opt->tot_len);
653 dst_release(dst);
654 return NULL;
657 /* The socket must have it's spinlock held when we get
658 * here.
660 * We have a potential double-lock case here, so even when
661 * doing backlog processing we use the BH locking scheme.
662 * This is because we cannot sleep with the original spinlock
663 * held.
665 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
667 struct ipv6_pinfo *np = inet6_sk(sk);
668 struct sk_buff *opt_skb = NULL;
670 /* Imagine: socket is IPv6. IPv4 packet arrives,
671 goes to IPv4 receive handler and backlogged.
672 From backlog it always goes here. Kerboom...
673 Fortunately, dccp_rcv_established and rcv_established
674 handle them correctly, but it is not case with
675 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
678 if (skb->protocol == htons(ETH_P_IP))
679 return dccp_v4_do_rcv(sk, skb);
681 if (sk_filter(sk, skb))
682 goto discard;
685 * socket locking is here for SMP purposes as backlog rcv is currently
686 * called with bh processing disabled.
689 /* Do Stevens' IPV6_PKTOPTIONS.
691 Yes, guys, it is the only place in our code, where we
692 may make it not affecting IPv4.
693 The rest of code is protocol independent,
694 and I do not like idea to uglify IPv4.
696 Actually, all the idea behind IPV6_PKTOPTIONS
697 looks not very well thought. For now we latch
698 options, received in the last packet, enqueued
699 by tcp. Feel free to propose better solution.
700 --ANK (980728)
702 if (np->rxopt.all)
704 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
705 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
707 opt_skb = skb_clone(skb, GFP_ATOMIC);
709 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
710 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
711 goto reset;
712 if (opt_skb) {
713 /* XXX This is where we would goto ipv6_pktoptions. */
714 __kfree_skb(opt_skb);
716 return 0;
720 * Step 3: Process LISTEN state
721 * If S.state == LISTEN,
722 * If P.type == Request or P contains a valid Init Cookie option,
723 * (* Must scan the packet's options to check for Init
724 * Cookies. Only Init Cookies are processed here,
725 * however; other options are processed in Step 8. This
726 * scan need only be performed if the endpoint uses Init
727 * Cookies *)
728 * (* Generate a new socket and switch to that socket *)
729 * Set S := new socket for this port pair
730 * S.state = RESPOND
731 * Choose S.ISS (initial seqno) or set from Init Cookies
732 * Initialize S.GAR := S.ISS
733 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
734 * Continue with S.state == RESPOND
735 * (* A Response packet will be generated in Step 11 *)
736 * Otherwise,
737 * Generate Reset(No Connection) unless P.type == Reset
738 * Drop packet and return
740 * NOTE: the check for the packet types is done in
741 * dccp_rcv_state_process
743 if (sk->sk_state == DCCP_LISTEN) {
744 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
746 if (nsk == NULL)
747 goto discard;
749 * Queue it on the new socket if the new socket is active,
750 * otherwise we just shortcircuit this and continue with
751 * the new socket..
753 if (nsk != sk) {
754 if (dccp_child_process(sk, nsk, skb))
755 goto reset;
756 if (opt_skb != NULL)
757 __kfree_skb(opt_skb);
758 return 0;
762 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
763 goto reset;
764 if (opt_skb) {
765 /* XXX This is where we would goto ipv6_pktoptions. */
766 __kfree_skb(opt_skb);
768 return 0;
770 reset:
771 dccp_v6_ctl_send_reset(sk, skb);
772 discard:
773 if (opt_skb != NULL)
774 __kfree_skb(opt_skb);
775 kfree_skb(skb);
776 return 0;
779 static int dccp_v6_rcv(struct sk_buff *skb)
781 const struct dccp_hdr *dh;
782 struct sock *sk;
783 int min_cov;
785 /* Step 1: Check header basics */
787 if (dccp_invalid_packet(skb))
788 goto discard_it;
790 /* Step 1: If header checksum is incorrect, drop packet and return. */
791 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
792 &ipv6_hdr(skb)->daddr)) {
793 DCCP_WARN("dropped packet with invalid checksum\n");
794 goto discard_it;
797 dh = dccp_hdr(skb);
799 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
800 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
802 if (dccp_packet_without_ack(skb))
803 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
804 else
805 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
807 /* Step 2:
808 * Look up flow ID in table and get corresponding socket */
809 sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
810 dh->dccph_sport, dh->dccph_dport);
812 * Step 2:
813 * If no socket ...
815 if (sk == NULL) {
816 dccp_pr_debug("failed to look up flow ID in table and "
817 "get corresponding socket\n");
818 goto no_dccp_socket;
822 * Step 2:
823 * ... or S.state == TIMEWAIT,
824 * Generate Reset(No Connection) unless P.type == Reset
825 * Drop packet and return
827 if (sk->sk_state == DCCP_TIME_WAIT) {
828 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
829 inet_twsk_put(inet_twsk(sk));
830 goto no_dccp_socket;
834 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
835 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
836 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
838 min_cov = dccp_sk(sk)->dccps_pcrlen;
839 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
840 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
841 dh->dccph_cscov, min_cov);
842 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
843 goto discard_and_relse;
846 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
847 goto discard_and_relse;
849 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
851 no_dccp_socket:
852 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
853 goto discard_it;
855 * Step 2:
856 * If no socket ...
857 * Generate Reset(No Connection) unless P.type == Reset
858 * Drop packet and return
860 if (dh->dccph_type != DCCP_PKT_RESET) {
861 DCCP_SKB_CB(skb)->dccpd_reset_code =
862 DCCP_RESET_CODE_NO_CONNECTION;
863 dccp_v6_ctl_send_reset(sk, skb);
866 discard_it:
867 kfree_skb(skb);
868 return 0;
870 discard_and_relse:
871 sock_put(sk);
872 goto discard_it;
875 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
876 int addr_len)
878 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
879 struct inet_connection_sock *icsk = inet_csk(sk);
880 struct inet_sock *inet = inet_sk(sk);
881 struct ipv6_pinfo *np = inet6_sk(sk);
882 struct dccp_sock *dp = dccp_sk(sk);
883 struct in6_addr *saddr = NULL, *final_p = NULL, final;
884 struct flowi fl;
885 struct dst_entry *dst;
886 int addr_type;
887 int err;
889 dp->dccps_role = DCCP_ROLE_CLIENT;
891 if (addr_len < SIN6_LEN_RFC2133)
892 return -EINVAL;
894 if (usin->sin6_family != AF_INET6)
895 return -EAFNOSUPPORT;
897 memset(&fl, 0, sizeof(fl));
899 if (np->sndflow) {
900 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
901 IP6_ECN_flow_init(fl.fl6_flowlabel);
902 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
903 struct ip6_flowlabel *flowlabel;
904 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
905 if (flowlabel == NULL)
906 return -EINVAL;
907 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
908 fl6_sock_release(flowlabel);
912 * connect() to INADDR_ANY means loopback (BSD'ism).
914 if (ipv6_addr_any(&usin->sin6_addr))
915 usin->sin6_addr.s6_addr[15] = 1;
917 addr_type = ipv6_addr_type(&usin->sin6_addr);
919 if (addr_type & IPV6_ADDR_MULTICAST)
920 return -ENETUNREACH;
922 if (addr_type & IPV6_ADDR_LINKLOCAL) {
923 if (addr_len >= sizeof(struct sockaddr_in6) &&
924 usin->sin6_scope_id) {
925 /* If interface is set while binding, indices
926 * must coincide.
928 if (sk->sk_bound_dev_if &&
929 sk->sk_bound_dev_if != usin->sin6_scope_id)
930 return -EINVAL;
932 sk->sk_bound_dev_if = usin->sin6_scope_id;
935 /* Connect to link-local address requires an interface */
936 if (!sk->sk_bound_dev_if)
937 return -EINVAL;
940 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
941 np->flow_label = fl.fl6_flowlabel;
944 * DCCP over IPv4
946 if (addr_type == IPV6_ADDR_MAPPED) {
947 u32 exthdrlen = icsk->icsk_ext_hdr_len;
948 struct sockaddr_in sin;
950 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
952 if (__ipv6_only_sock(sk))
953 return -ENETUNREACH;
955 sin.sin_family = AF_INET;
956 sin.sin_port = usin->sin6_port;
957 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
959 icsk->icsk_af_ops = &dccp_ipv6_mapped;
960 sk->sk_backlog_rcv = dccp_v4_do_rcv;
962 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
963 if (err) {
964 icsk->icsk_ext_hdr_len = exthdrlen;
965 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
966 sk->sk_backlog_rcv = dccp_v6_do_rcv;
967 goto failure;
968 } else {
969 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
970 inet->saddr);
971 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
972 inet->rcv_saddr);
975 return err;
978 if (!ipv6_addr_any(&np->rcv_saddr))
979 saddr = &np->rcv_saddr;
981 fl.proto = IPPROTO_DCCP;
982 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
983 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
984 fl.oif = sk->sk_bound_dev_if;
985 fl.fl_ip_dport = usin->sin6_port;
986 fl.fl_ip_sport = inet->sport;
987 security_sk_classify_flow(sk, &fl);
989 if (np->opt != NULL && np->opt->srcrt != NULL) {
990 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
992 ipv6_addr_copy(&final, &fl.fl6_dst);
993 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
994 final_p = &final;
997 err = ip6_dst_lookup(sk, &dst, &fl);
998 if (err)
999 goto failure;
1001 if (final_p)
1002 ipv6_addr_copy(&fl.fl6_dst, final_p);
1004 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
1005 if (err < 0) {
1006 if (err == -EREMOTE)
1007 err = ip6_dst_blackhole(sk, &dst, &fl);
1008 if (err < 0)
1009 goto failure;
1012 if (saddr == NULL) {
1013 saddr = &fl.fl6_src;
1014 ipv6_addr_copy(&np->rcv_saddr, saddr);
1017 /* set the source address */
1018 ipv6_addr_copy(&np->saddr, saddr);
1019 inet->rcv_saddr = LOOPBACK4_IPV6;
1021 __ip6_dst_store(sk, dst, NULL, NULL);
1023 icsk->icsk_ext_hdr_len = 0;
1024 if (np->opt != NULL)
1025 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1026 np->opt->opt_nflen);
1028 inet->dport = usin->sin6_port;
1030 dccp_set_state(sk, DCCP_REQUESTING);
1031 err = inet6_hash_connect(&dccp_death_row, sk);
1032 if (err)
1033 goto late_failure;
1035 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1036 np->daddr.s6_addr32,
1037 inet->sport, inet->dport);
1038 err = dccp_connect(sk);
1039 if (err)
1040 goto late_failure;
1042 return 0;
1044 late_failure:
1045 dccp_set_state(sk, DCCP_CLOSED);
1046 __sk_dst_reset(sk);
1047 failure:
1048 inet->dport = 0;
1049 sk->sk_route_caps = 0;
1050 return err;
1053 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1054 .queue_xmit = inet6_csk_xmit,
1055 .send_check = dccp_v6_send_check,
1056 .rebuild_header = inet6_sk_rebuild_header,
1057 .conn_request = dccp_v6_conn_request,
1058 .syn_recv_sock = dccp_v6_request_recv_sock,
1059 .net_header_len = sizeof(struct ipv6hdr),
1060 .setsockopt = ipv6_setsockopt,
1061 .getsockopt = ipv6_getsockopt,
1062 .addr2sockaddr = inet6_csk_addr2sockaddr,
1063 .sockaddr_len = sizeof(struct sockaddr_in6),
1064 .bind_conflict = inet6_csk_bind_conflict,
1065 #ifdef CONFIG_COMPAT
1066 .compat_setsockopt = compat_ipv6_setsockopt,
1067 .compat_getsockopt = compat_ipv6_getsockopt,
1068 #endif
1072 * DCCP over IPv4 via INET6 API
1074 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1075 .queue_xmit = ip_queue_xmit,
1076 .send_check = dccp_v4_send_check,
1077 .rebuild_header = inet_sk_rebuild_header,
1078 .conn_request = dccp_v6_conn_request,
1079 .syn_recv_sock = dccp_v6_request_recv_sock,
1080 .net_header_len = sizeof(struct iphdr),
1081 .setsockopt = ipv6_setsockopt,
1082 .getsockopt = ipv6_getsockopt,
1083 .addr2sockaddr = inet6_csk_addr2sockaddr,
1084 .sockaddr_len = sizeof(struct sockaddr_in6),
1085 #ifdef CONFIG_COMPAT
1086 .compat_setsockopt = compat_ipv6_setsockopt,
1087 .compat_getsockopt = compat_ipv6_getsockopt,
1088 #endif
1091 /* NOTE: A lot of things set to zero explicitly by call to
1092 * sk_alloc() so need not be done here.
1094 static int dccp_v6_init_sock(struct sock *sk)
1096 static __u8 dccp_v6_ctl_sock_initialized;
1097 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1099 if (err == 0) {
1100 if (unlikely(!dccp_v6_ctl_sock_initialized))
1101 dccp_v6_ctl_sock_initialized = 1;
1102 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1105 return err;
1108 static void dccp_v6_destroy_sock(struct sock *sk)
1110 dccp_destroy_sock(sk);
1111 inet6_destroy_sock(sk);
1114 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1115 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1118 static struct proto dccp_v6_prot = {
1119 .name = "DCCPv6",
1120 .owner = THIS_MODULE,
1121 .close = dccp_close,
1122 .connect = dccp_v6_connect,
1123 .disconnect = dccp_disconnect,
1124 .ioctl = dccp_ioctl,
1125 .init = dccp_v6_init_sock,
1126 .setsockopt = dccp_setsockopt,
1127 .getsockopt = dccp_getsockopt,
1128 .sendmsg = dccp_sendmsg,
1129 .recvmsg = dccp_recvmsg,
1130 .backlog_rcv = dccp_v6_do_rcv,
1131 .hash = dccp_v6_hash,
1132 .unhash = inet_unhash,
1133 .accept = inet_csk_accept,
1134 .get_port = inet_csk_get_port,
1135 .shutdown = dccp_shutdown,
1136 .destroy = dccp_v6_destroy_sock,
1137 .orphan_count = &dccp_orphan_count,
1138 .max_header = MAX_DCCP_HEADER,
1139 .obj_size = sizeof(struct dccp6_sock),
1140 .slab_flags = SLAB_DESTROY_BY_RCU,
1141 .rsk_prot = &dccp6_request_sock_ops,
1142 .twsk_prot = &dccp6_timewait_sock_ops,
1143 .h.hashinfo = &dccp_hashinfo,
1144 #ifdef CONFIG_COMPAT
1145 .compat_setsockopt = compat_dccp_setsockopt,
1146 .compat_getsockopt = compat_dccp_getsockopt,
1147 #endif
1150 static const struct inet6_protocol dccp_v6_protocol = {
1151 .handler = dccp_v6_rcv,
1152 .err_handler = dccp_v6_err,
1153 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1156 static const struct proto_ops inet6_dccp_ops = {
1157 .family = PF_INET6,
1158 .owner = THIS_MODULE,
1159 .release = inet6_release,
1160 .bind = inet6_bind,
1161 .connect = inet_stream_connect,
1162 .socketpair = sock_no_socketpair,
1163 .accept = inet_accept,
1164 .getname = inet6_getname,
1165 .poll = dccp_poll,
1166 .ioctl = inet6_ioctl,
1167 .listen = inet_dccp_listen,
1168 .shutdown = inet_shutdown,
1169 .setsockopt = sock_common_setsockopt,
1170 .getsockopt = sock_common_getsockopt,
1171 .sendmsg = inet_sendmsg,
1172 .recvmsg = sock_common_recvmsg,
1173 .mmap = sock_no_mmap,
1174 .sendpage = sock_no_sendpage,
1175 #ifdef CONFIG_COMPAT
1176 .compat_setsockopt = compat_sock_common_setsockopt,
1177 .compat_getsockopt = compat_sock_common_getsockopt,
1178 #endif
1181 static struct inet_protosw dccp_v6_protosw = {
1182 .type = SOCK_DCCP,
1183 .protocol = IPPROTO_DCCP,
1184 .prot = &dccp_v6_prot,
1185 .ops = &inet6_dccp_ops,
1186 .capability = -1,
1187 .flags = INET_PROTOSW_ICSK,
1190 static int dccp_v6_init_net(struct net *net)
1192 int err;
1194 err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1195 SOCK_DCCP, IPPROTO_DCCP, net);
1196 return err;
1199 static void dccp_v6_exit_net(struct net *net)
1201 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1204 static struct pernet_operations dccp_v6_ops = {
1205 .init = dccp_v6_init_net,
1206 .exit = dccp_v6_exit_net,
1209 static int __init dccp_v6_init(void)
1211 int err = proto_register(&dccp_v6_prot, 1);
1213 if (err != 0)
1214 goto out;
1216 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1217 if (err != 0)
1218 goto out_unregister_proto;
1220 inet6_register_protosw(&dccp_v6_protosw);
1222 err = register_pernet_subsys(&dccp_v6_ops);
1223 if (err != 0)
1224 goto out_destroy_ctl_sock;
1225 out:
1226 return err;
1228 out_destroy_ctl_sock:
1229 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1230 inet6_unregister_protosw(&dccp_v6_protosw);
1231 out_unregister_proto:
1232 proto_unregister(&dccp_v6_prot);
1233 goto out;
1236 static void __exit dccp_v6_exit(void)
1238 unregister_pernet_subsys(&dccp_v6_ops);
1239 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1240 inet6_unregister_protosw(&dccp_v6_protosw);
1241 proto_unregister(&dccp_v6_prot);
1244 module_init(dccp_v6_init);
1245 module_exit(dccp_v6_exit);
1248 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1249 * values directly, Also cover the case where the protocol is not specified,
1250 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1252 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1253 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1254 MODULE_LICENSE("GPL");
1255 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1256 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");