net: Pass optional SKB and SK arguments to dst_ops->{update_pmtu,redirect}()
[linux-3.10.git] / net / dccp / ipv6.c
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 == NDISC_REDIRECT) {
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
140         if (type == ICMPV6_PKT_TOOBIG) {
141                 struct dst_entry *dst = NULL;
142
143                 if (sock_owned_by_user(sk))
144                         goto out;
145                 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
146                         goto out;
147
148                 dst = inet6_csk_update_pmtu(sk, ntohl(info));
149                 if (!dst)
150                         goto out;
151
152                 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
153                         dccp_sync_mss(sk, dst_mtu(dst));
154                 goto out;
155         }
156
157         icmpv6_err_convert(type, code, &err);
158
159         /* Might be for an request_sock */
160         switch (sk->sk_state) {
161                 struct request_sock *req, **prev;
162         case DCCP_LISTEN:
163                 if (sock_owned_by_user(sk))
164                         goto out;
165
166                 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
167                                            &hdr->daddr, &hdr->saddr,
168                                            inet6_iif(skb));
169                 if (req == NULL)
170                         goto out;
171
172                 /*
173                  * ICMPs are not backlogged, hence we cannot get an established
174                  * socket here.
175                  */
176                 WARN_ON(req->sk != NULL);
177
178                 if (!between48(seq, dccp_rsk(req)->dreq_iss,
179                                     dccp_rsk(req)->dreq_gss)) {
180                         NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
181                         goto out;
182                 }
183
184                 inet_csk_reqsk_queue_drop(sk, req, prev);
185                 goto out;
186
187         case DCCP_REQUESTING:
188         case DCCP_RESPOND:  /* Cannot happen.
189                                It can, it SYNs are crossed. --ANK */
190                 if (!sock_owned_by_user(sk)) {
191                         DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
192                         sk->sk_err = err;
193                         /*
194                          * Wake people up to see the error
195                          * (see connect in sock.c)
196                          */
197                         sk->sk_error_report(sk);
198                         dccp_done(sk);
199                 } else
200                         sk->sk_err_soft = err;
201                 goto out;
202         }
203
204         if (!sock_owned_by_user(sk) && np->recverr) {
205                 sk->sk_err = err;
206                 sk->sk_error_report(sk);
207         } else
208                 sk->sk_err_soft = err;
209
210 out:
211         bh_unlock_sock(sk);
212         sock_put(sk);
213 }
214
215
216 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
217                                  struct request_values *rv_unused)
218 {
219         struct inet6_request_sock *ireq6 = inet6_rsk(req);
220         struct ipv6_pinfo *np = inet6_sk(sk);
221         struct sk_buff *skb;
222         struct in6_addr *final_p, final;
223         struct flowi6 fl6;
224         int err = -1;
225         struct dst_entry *dst;
226
227         memset(&fl6, 0, sizeof(fl6));
228         fl6.flowi6_proto = IPPROTO_DCCP;
229         fl6.daddr = ireq6->rmt_addr;
230         fl6.saddr = ireq6->loc_addr;
231         fl6.flowlabel = 0;
232         fl6.flowi6_oif = ireq6->iif;
233         fl6.fl6_dport = inet_rsk(req)->rmt_port;
234         fl6.fl6_sport = inet_rsk(req)->loc_port;
235         security_req_classify_flow(req, flowi6_to_flowi(&fl6));
236
237
238         final_p = fl6_update_dst(&fl6, np->opt, &final);
239
240         dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
241         if (IS_ERR(dst)) {
242                 err = PTR_ERR(dst);
243                 dst = NULL;
244                 goto done;
245         }
246
247         skb = dccp_make_response(sk, dst, req);
248         if (skb != NULL) {
249                 struct dccp_hdr *dh = dccp_hdr(skb);
250
251                 dh->dccph_checksum = dccp_v6_csum_finish(skb,
252                                                          &ireq6->loc_addr,
253                                                          &ireq6->rmt_addr);
254                 fl6.daddr = ireq6->rmt_addr;
255                 err = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass);
256                 err = net_xmit_eval(err);
257         }
258
259 done:
260         dst_release(dst);
261         return err;
262 }
263
264 static void dccp_v6_reqsk_destructor(struct request_sock *req)
265 {
266         dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
267         if (inet6_rsk(req)->pktopts != NULL)
268                 kfree_skb(inet6_rsk(req)->pktopts);
269 }
270
271 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
272 {
273         const struct ipv6hdr *rxip6h;
274         struct sk_buff *skb;
275         struct flowi6 fl6;
276         struct net *net = dev_net(skb_dst(rxskb)->dev);
277         struct sock *ctl_sk = net->dccp.v6_ctl_sk;
278         struct dst_entry *dst;
279
280         if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
281                 return;
282
283         if (!ipv6_unicast_destination(rxskb))
284                 return;
285
286         skb = dccp_ctl_make_reset(ctl_sk, rxskb);
287         if (skb == NULL)
288                 return;
289
290         rxip6h = ipv6_hdr(rxskb);
291         dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
292                                                             &rxip6h->daddr);
293
294         memset(&fl6, 0, sizeof(fl6));
295         fl6.daddr = rxip6h->saddr;
296         fl6.saddr = rxip6h->daddr;
297
298         fl6.flowi6_proto = IPPROTO_DCCP;
299         fl6.flowi6_oif = inet6_iif(rxskb);
300         fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
301         fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
302         security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
303
304         /* sk = NULL, but it is safe for now. RST socket required. */
305         dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
306         if (!IS_ERR(dst)) {
307                 skb_dst_set(skb, dst);
308                 ip6_xmit(ctl_sk, skb, &fl6, NULL, 0);
309                 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
310                 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
311                 return;
312         }
313
314         kfree_skb(skb);
315 }
316
317 static struct request_sock_ops dccp6_request_sock_ops = {
318         .family         = AF_INET6,
319         .obj_size       = sizeof(struct dccp6_request_sock),
320         .rtx_syn_ack    = dccp_v6_send_response,
321         .send_ack       = dccp_reqsk_send_ack,
322         .destructor     = dccp_v6_reqsk_destructor,
323         .send_reset     = dccp_v6_ctl_send_reset,
324         .syn_ack_timeout = dccp_syn_ack_timeout,
325 };
326
327 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
328 {
329         const struct dccp_hdr *dh = dccp_hdr(skb);
330         const struct ipv6hdr *iph = ipv6_hdr(skb);
331         struct sock *nsk;
332         struct request_sock **prev;
333         /* Find possible connection requests. */
334         struct request_sock *req = inet6_csk_search_req(sk, &prev,
335                                                         dh->dccph_sport,
336                                                         &iph->saddr,
337                                                         &iph->daddr,
338                                                         inet6_iif(skb));
339         if (req != NULL)
340                 return dccp_check_req(sk, skb, req, prev);
341
342         nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
343                                          &iph->saddr, dh->dccph_sport,
344                                          &iph->daddr, ntohs(dh->dccph_dport),
345                                          inet6_iif(skb));
346         if (nsk != NULL) {
347                 if (nsk->sk_state != DCCP_TIME_WAIT) {
348                         bh_lock_sock(nsk);
349                         return nsk;
350                 }
351                 inet_twsk_put(inet_twsk(nsk));
352                 return NULL;
353         }
354
355         return sk;
356 }
357
358 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
359 {
360         struct request_sock *req;
361         struct dccp_request_sock *dreq;
362         struct inet6_request_sock *ireq6;
363         struct ipv6_pinfo *np = inet6_sk(sk);
364         const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
365         struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
366
367         if (skb->protocol == htons(ETH_P_IP))
368                 return dccp_v4_conn_request(sk, skb);
369
370         if (!ipv6_unicast_destination(skb))
371                 return 0;       /* discard, don't send a reset here */
372
373         if (dccp_bad_service_code(sk, service)) {
374                 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
375                 goto drop;
376         }
377         /*
378          * There are no SYN attacks on IPv6, yet...
379          */
380         dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
381         if (inet_csk_reqsk_queue_is_full(sk))
382                 goto drop;
383
384         if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
385                 goto drop;
386
387         req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
388         if (req == NULL)
389                 goto drop;
390
391         if (dccp_reqsk_init(req, dccp_sk(sk), skb))
392                 goto drop_and_free;
393
394         dreq = dccp_rsk(req);
395         if (dccp_parse_options(sk, dreq, skb))
396                 goto drop_and_free;
397
398         if (security_inet_conn_request(sk, skb, req))
399                 goto drop_and_free;
400
401         ireq6 = inet6_rsk(req);
402         ireq6->rmt_addr = ipv6_hdr(skb)->saddr;
403         ireq6->loc_addr = ipv6_hdr(skb)->daddr;
404
405         if (ipv6_opt_accepted(sk, skb) ||
406             np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
407             np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
408                 atomic_inc(&skb->users);
409                 ireq6->pktopts = skb;
410         }
411         ireq6->iif = sk->sk_bound_dev_if;
412
413         /* So that link locals have meaning */
414         if (!sk->sk_bound_dev_if &&
415             ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
416                 ireq6->iif = inet6_iif(skb);
417
418         /*
419          * Step 3: Process LISTEN state
420          *
421          *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
422          *
423          * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
424          */
425         dreq->dreq_isr     = dcb->dccpd_seq;
426         dreq->dreq_gsr     = dreq->dreq_isr;
427         dreq->dreq_iss     = dccp_v6_init_sequence(skb);
428         dreq->dreq_gss     = dreq->dreq_iss;
429         dreq->dreq_service = service;
430
431         if (dccp_v6_send_response(sk, req, NULL))
432                 goto drop_and_free;
433
434         inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
435         return 0;
436
437 drop_and_free:
438         reqsk_free(req);
439 drop:
440         DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
441         return -1;
442 }
443
444 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
445                                               struct sk_buff *skb,
446                                               struct request_sock *req,
447                                               struct dst_entry *dst)
448 {
449         struct inet6_request_sock *ireq6 = inet6_rsk(req);
450         struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
451         struct inet_sock *newinet;
452         struct dccp6_sock *newdp6;
453         struct sock *newsk;
454
455         if (skb->protocol == htons(ETH_P_IP)) {
456                 /*
457                  *      v6 mapped
458                  */
459                 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
460                 if (newsk == NULL)
461                         return NULL;
462
463                 newdp6 = (struct dccp6_sock *)newsk;
464                 newinet = inet_sk(newsk);
465                 newinet->pinet6 = &newdp6->inet6;
466                 newnp = inet6_sk(newsk);
467
468                 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
469
470                 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
471
472                 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
473
474                 newnp->rcv_saddr = newnp->saddr;
475
476                 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
477                 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
478                 newnp->pktoptions  = NULL;
479                 newnp->opt         = NULL;
480                 newnp->mcast_oif   = inet6_iif(skb);
481                 newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
482
483                 /*
484                  * No need to charge this sock to the relevant IPv6 refcnt debug socks count
485                  * here, dccp_create_openreq_child now does this for us, see the comment in
486                  * that function for the gory details. -acme
487                  */
488
489                 /* It is tricky place. Until this moment IPv4 tcp
490                    worked with IPv6 icsk.icsk_af_ops.
491                    Sync it now.
492                  */
493                 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
494
495                 return newsk;
496         }
497
498
499         if (sk_acceptq_is_full(sk))
500                 goto out_overflow;
501
502         if (dst == NULL) {
503                 struct in6_addr *final_p, final;
504                 struct flowi6 fl6;
505
506                 memset(&fl6, 0, sizeof(fl6));
507                 fl6.flowi6_proto = IPPROTO_DCCP;
508                 fl6.daddr = ireq6->rmt_addr;
509                 final_p = fl6_update_dst(&fl6, np->opt, &final);
510                 fl6.saddr = ireq6->loc_addr;
511                 fl6.flowi6_oif = sk->sk_bound_dev_if;
512                 fl6.fl6_dport = inet_rsk(req)->rmt_port;
513                 fl6.fl6_sport = inet_rsk(req)->loc_port;
514                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
515
516                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
517                 if (IS_ERR(dst))
518                         goto out;
519         }
520
521         newsk = dccp_create_openreq_child(sk, req, skb);
522         if (newsk == NULL)
523                 goto out_nonewsk;
524
525         /*
526          * No need to charge this sock to the relevant IPv6 refcnt debug socks
527          * count here, dccp_create_openreq_child now does this for us, see the
528          * comment in that function for the gory details. -acme
529          */
530
531         __ip6_dst_store(newsk, dst, NULL, NULL);
532         newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
533                                                       NETIF_F_TSO);
534         newdp6 = (struct dccp6_sock *)newsk;
535         newinet = inet_sk(newsk);
536         newinet->pinet6 = &newdp6->inet6;
537         newnp = inet6_sk(newsk);
538
539         memcpy(newnp, np, sizeof(struct ipv6_pinfo));
540
541         newnp->daddr = ireq6->rmt_addr;
542         newnp->saddr = ireq6->loc_addr;
543         newnp->rcv_saddr = ireq6->loc_addr;
544         newsk->sk_bound_dev_if = ireq6->iif;
545
546         /* Now IPv6 options...
547
548            First: no IPv4 options.
549          */
550         newinet->inet_opt = NULL;
551
552         /* Clone RX bits */
553         newnp->rxopt.all = np->rxopt.all;
554
555         /* Clone pktoptions received with SYN */
556         newnp->pktoptions = NULL;
557         if (ireq6->pktopts != NULL) {
558                 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
559                 consume_skb(ireq6->pktopts);
560                 ireq6->pktopts = NULL;
561                 if (newnp->pktoptions)
562                         skb_set_owner_r(newnp->pktoptions, newsk);
563         }
564         newnp->opt        = NULL;
565         newnp->mcast_oif  = inet6_iif(skb);
566         newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
567
568         /*
569          * Clone native IPv6 options from listening socket (if any)
570          *
571          * Yes, keeping reference count would be much more clever, but we make
572          * one more one thing there: reattach optmem to newsk.
573          */
574         if (np->opt != NULL)
575                 newnp->opt = ipv6_dup_options(newsk, np->opt);
576
577         inet_csk(newsk)->icsk_ext_hdr_len = 0;
578         if (newnp->opt != NULL)
579                 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
580                                                      newnp->opt->opt_flen);
581
582         dccp_sync_mss(newsk, dst_mtu(dst));
583
584         newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
585         newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
586
587         if (__inet_inherit_port(sk, newsk) < 0) {
588                 sock_put(newsk);
589                 goto out;
590         }
591         __inet6_hash(newsk, NULL);
592
593         return newsk;
594
595 out_overflow:
596         NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
597 out_nonewsk:
598         dst_release(dst);
599 out:
600         NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
601         return NULL;
602 }
603
604 /* The socket must have it's spinlock held when we get
605  * here.
606  *
607  * We have a potential double-lock case here, so even when
608  * doing backlog processing we use the BH locking scheme.
609  * This is because we cannot sleep with the original spinlock
610  * held.
611  */
612 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
613 {
614         struct ipv6_pinfo *np = inet6_sk(sk);
615         struct sk_buff *opt_skb = NULL;
616
617         /* Imagine: socket is IPv6. IPv4 packet arrives,
618            goes to IPv4 receive handler and backlogged.
619            From backlog it always goes here. Kerboom...
620            Fortunately, dccp_rcv_established and rcv_established
621            handle them correctly, but it is not case with
622            dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
623          */
624
625         if (skb->protocol == htons(ETH_P_IP))
626                 return dccp_v4_do_rcv(sk, skb);
627
628         if (sk_filter(sk, skb))
629                 goto discard;
630
631         /*
632          * socket locking is here for SMP purposes as backlog rcv is currently
633          * called with bh processing disabled.
634          */
635
636         /* Do Stevens' IPV6_PKTOPTIONS.
637
638            Yes, guys, it is the only place in our code, where we
639            may make it not affecting IPv4.
640            The rest of code is protocol independent,
641            and I do not like idea to uglify IPv4.
642
643            Actually, all the idea behind IPV6_PKTOPTIONS
644            looks not very well thought. For now we latch
645            options, received in the last packet, enqueued
646            by tcp. Feel free to propose better solution.
647                                                --ANK (980728)
648          */
649         if (np->rxopt.all)
650         /*
651          * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
652          *        (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
653          */
654                 opt_skb = skb_clone(skb, GFP_ATOMIC);
655
656         if (sk->sk_state == DCCP_OPEN) { /* Fast path */
657                 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
658                         goto reset;
659                 if (opt_skb) {
660                         /* XXX This is where we would goto ipv6_pktoptions. */
661                         __kfree_skb(opt_skb);
662                 }
663                 return 0;
664         }
665
666         /*
667          *  Step 3: Process LISTEN state
668          *     If S.state == LISTEN,
669          *       If P.type == Request or P contains a valid Init Cookie option,
670          *            (* Must scan the packet's options to check for Init
671          *               Cookies.  Only Init Cookies are processed here,
672          *               however; other options are processed in Step 8.  This
673          *               scan need only be performed if the endpoint uses Init
674          *               Cookies *)
675          *            (* Generate a new socket and switch to that socket *)
676          *            Set S := new socket for this port pair
677          *            S.state = RESPOND
678          *            Choose S.ISS (initial seqno) or set from Init Cookies
679          *            Initialize S.GAR := S.ISS
680          *            Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
681          *            Continue with S.state == RESPOND
682          *            (* A Response packet will be generated in Step 11 *)
683          *       Otherwise,
684          *            Generate Reset(No Connection) unless P.type == Reset
685          *            Drop packet and return
686          *
687          * NOTE: the check for the packet types is done in
688          *       dccp_rcv_state_process
689          */
690         if (sk->sk_state == DCCP_LISTEN) {
691                 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
692
693                 if (nsk == NULL)
694                         goto discard;
695                 /*
696                  * Queue it on the new socket if the new socket is active,
697                  * otherwise we just shortcircuit this and continue with
698                  * the new socket..
699                  */
700                 if (nsk != sk) {
701                         if (dccp_child_process(sk, nsk, skb))
702                                 goto reset;
703                         if (opt_skb != NULL)
704                                 __kfree_skb(opt_skb);
705                         return 0;
706                 }
707         }
708
709         if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
710                 goto reset;
711         if (opt_skb) {
712                 /* XXX This is where we would goto ipv6_pktoptions. */
713                 __kfree_skb(opt_skb);
714         }
715         return 0;
716
717 reset:
718         dccp_v6_ctl_send_reset(sk, skb);
719 discard:
720         if (opt_skb != NULL)
721                 __kfree_skb(opt_skb);
722         kfree_skb(skb);
723         return 0;
724 }
725
726 static int dccp_v6_rcv(struct sk_buff *skb)
727 {
728         const struct dccp_hdr *dh;
729         struct sock *sk;
730         int min_cov;
731
732         /* Step 1: Check header basics */
733
734         if (dccp_invalid_packet(skb))
735                 goto discard_it;
736
737         /* Step 1: If header checksum is incorrect, drop packet and return. */
738         if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
739                                      &ipv6_hdr(skb)->daddr)) {
740                 DCCP_WARN("dropped packet with invalid checksum\n");
741                 goto discard_it;
742         }
743
744         dh = dccp_hdr(skb);
745
746         DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
747         DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
748
749         if (dccp_packet_without_ack(skb))
750                 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
751         else
752                 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
753
754         /* Step 2:
755          *      Look up flow ID in table and get corresponding socket */
756         sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
757                                 dh->dccph_sport, dh->dccph_dport);
758         /*
759          * Step 2:
760          *      If no socket ...
761          */
762         if (sk == NULL) {
763                 dccp_pr_debug("failed to look up flow ID in table and "
764                               "get corresponding socket\n");
765                 goto no_dccp_socket;
766         }
767
768         /*
769          * Step 2:
770          *      ... or S.state == TIMEWAIT,
771          *              Generate Reset(No Connection) unless P.type == Reset
772          *              Drop packet and return
773          */
774         if (sk->sk_state == DCCP_TIME_WAIT) {
775                 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
776                 inet_twsk_put(inet_twsk(sk));
777                 goto no_dccp_socket;
778         }
779
780         /*
781          * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
782          *      o if MinCsCov = 0, only packets with CsCov = 0 are accepted
783          *      o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
784          */
785         min_cov = dccp_sk(sk)->dccps_pcrlen;
786         if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
787                 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
788                               dh->dccph_cscov, min_cov);
789                 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
790                 goto discard_and_relse;
791         }
792
793         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
794                 goto discard_and_relse;
795
796         return sk_receive_skb(sk, skb, 1) ? -1 : 0;
797
798 no_dccp_socket:
799         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
800                 goto discard_it;
801         /*
802          * Step 2:
803          *      If no socket ...
804          *              Generate Reset(No Connection) unless P.type == Reset
805          *              Drop packet and return
806          */
807         if (dh->dccph_type != DCCP_PKT_RESET) {
808                 DCCP_SKB_CB(skb)->dccpd_reset_code =
809                                         DCCP_RESET_CODE_NO_CONNECTION;
810                 dccp_v6_ctl_send_reset(sk, skb);
811         }
812
813 discard_it:
814         kfree_skb(skb);
815         return 0;
816
817 discard_and_relse:
818         sock_put(sk);
819         goto discard_it;
820 }
821
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 flowi6 fl6;
832         struct dst_entry *dst;
833         int addr_type;
834         int err;
835
836         dp->dccps_role = DCCP_ROLE_CLIENT;
837
838         if (addr_len < SIN6_LEN_RFC2133)
839                 return -EINVAL;
840
841         if (usin->sin6_family != AF_INET6)
842                 return -EAFNOSUPPORT;
843
844         memset(&fl6, 0, sizeof(fl6));
845
846         if (np->sndflow) {
847                 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
848                 IP6_ECN_flow_init(fl6.flowlabel);
849                 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
850                         struct ip6_flowlabel *flowlabel;
851                         flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
852                         if (flowlabel == NULL)
853                                 return -EINVAL;
854                         usin->sin6_addr = flowlabel->dst;
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         np->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                 SOCK_DEBUG(sk, "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                 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
917                 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr);
918
919                 return err;
920         }
921
922         if (!ipv6_addr_any(&np->rcv_saddr))
923                 saddr = &np->rcv_saddr;
924
925         fl6.flowi6_proto = IPPROTO_DCCP;
926         fl6.daddr = np->daddr;
927         fl6.saddr = saddr ? *saddr : np->saddr;
928         fl6.flowi6_oif = sk->sk_bound_dev_if;
929         fl6.fl6_dport = usin->sin6_port;
930         fl6.fl6_sport = inet->inet_sport;
931         security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
932
933         final_p = fl6_update_dst(&fl6, np->opt, &final);
934
935         dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
936         if (IS_ERR(dst)) {
937                 err = PTR_ERR(dst);
938                 goto failure;
939         }
940
941         if (saddr == NULL) {
942                 saddr = &fl6.saddr;
943                 np->rcv_saddr = *saddr;
944         }
945
946         /* set the source address */
947         np->saddr = *saddr;
948         inet->inet_rcv_saddr = LOOPBACK4_IPV6;
949
950         __ip6_dst_store(sk, dst, NULL, NULL);
951
952         icsk->icsk_ext_hdr_len = 0;
953         if (np->opt != NULL)
954                 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
955                                           np->opt->opt_nflen);
956
957         inet->inet_dport = usin->sin6_port;
958
959         dccp_set_state(sk, DCCP_REQUESTING);
960         err = inet6_hash_connect(&dccp_death_row, sk);
961         if (err)
962                 goto late_failure;
963
964         dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
965                                                       np->daddr.s6_addr32,
966                                                       inet->inet_sport,
967                                                       inet->inet_dport);
968         err = dccp_connect(sk);
969         if (err)
970                 goto late_failure;
971
972         return 0;
973
974 late_failure:
975         dccp_set_state(sk, DCCP_CLOSED);
976         __sk_dst_reset(sk);
977 failure:
978         inet->inet_dport = 0;
979         sk->sk_route_caps = 0;
980         return err;
981 }
982
983 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
984         .queue_xmit        = inet6_csk_xmit,
985         .send_check        = dccp_v6_send_check,
986         .rebuild_header    = inet6_sk_rebuild_header,
987         .conn_request      = dccp_v6_conn_request,
988         .syn_recv_sock     = dccp_v6_request_recv_sock,
989         .net_header_len    = sizeof(struct ipv6hdr),
990         .setsockopt        = ipv6_setsockopt,
991         .getsockopt        = ipv6_getsockopt,
992         .addr2sockaddr     = inet6_csk_addr2sockaddr,
993         .sockaddr_len      = sizeof(struct sockaddr_in6),
994         .bind_conflict     = inet6_csk_bind_conflict,
995 #ifdef CONFIG_COMPAT
996         .compat_setsockopt = compat_ipv6_setsockopt,
997         .compat_getsockopt = compat_ipv6_getsockopt,
998 #endif
999 };
1000
1001 /*
1002  *      DCCP over IPv4 via INET6 API
1003  */
1004 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1005         .queue_xmit        = ip_queue_xmit,
1006         .send_check        = dccp_v4_send_check,
1007         .rebuild_header    = inet_sk_rebuild_header,
1008         .conn_request      = dccp_v6_conn_request,
1009         .syn_recv_sock     = dccp_v6_request_recv_sock,
1010         .net_header_len    = sizeof(struct iphdr),
1011         .setsockopt        = ipv6_setsockopt,
1012         .getsockopt        = ipv6_getsockopt,
1013         .addr2sockaddr     = inet6_csk_addr2sockaddr,
1014         .sockaddr_len      = sizeof(struct sockaddr_in6),
1015 #ifdef CONFIG_COMPAT
1016         .compat_setsockopt = compat_ipv6_setsockopt,
1017         .compat_getsockopt = compat_ipv6_getsockopt,
1018 #endif
1019 };
1020
1021 /* NOTE: A lot of things set to zero explicitly by call to
1022  *       sk_alloc() so need not be done here.
1023  */
1024 static int dccp_v6_init_sock(struct sock *sk)
1025 {
1026         static __u8 dccp_v6_ctl_sock_initialized;
1027         int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1028
1029         if (err == 0) {
1030                 if (unlikely(!dccp_v6_ctl_sock_initialized))
1031                         dccp_v6_ctl_sock_initialized = 1;
1032                 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1033         }
1034
1035         return err;
1036 }
1037
1038 static void dccp_v6_destroy_sock(struct sock *sk)
1039 {
1040         dccp_destroy_sock(sk);
1041         inet6_destroy_sock(sk);
1042 }
1043
1044 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1045         .twsk_obj_size  = sizeof(struct dccp6_timewait_sock),
1046 };
1047
1048 static struct proto dccp_v6_prot = {
1049         .name              = "DCCPv6",
1050         .owner             = THIS_MODULE,
1051         .close             = dccp_close,
1052         .connect           = dccp_v6_connect,
1053         .disconnect        = dccp_disconnect,
1054         .ioctl             = dccp_ioctl,
1055         .init              = dccp_v6_init_sock,
1056         .setsockopt        = dccp_setsockopt,
1057         .getsockopt        = dccp_getsockopt,
1058         .sendmsg           = dccp_sendmsg,
1059         .recvmsg           = dccp_recvmsg,
1060         .backlog_rcv       = dccp_v6_do_rcv,
1061         .hash              = dccp_v6_hash,
1062         .unhash            = inet_unhash,
1063         .accept            = inet_csk_accept,
1064         .get_port          = inet_csk_get_port,
1065         .shutdown          = dccp_shutdown,
1066         .destroy           = dccp_v6_destroy_sock,
1067         .orphan_count      = &dccp_orphan_count,
1068         .max_header        = MAX_DCCP_HEADER,
1069         .obj_size          = sizeof(struct dccp6_sock),
1070         .slab_flags        = SLAB_DESTROY_BY_RCU,
1071         .rsk_prot          = &dccp6_request_sock_ops,
1072         .twsk_prot         = &dccp6_timewait_sock_ops,
1073         .h.hashinfo        = &dccp_hashinfo,
1074 #ifdef CONFIG_COMPAT
1075         .compat_setsockopt = compat_dccp_setsockopt,
1076         .compat_getsockopt = compat_dccp_getsockopt,
1077 #endif
1078 };
1079
1080 static const struct inet6_protocol dccp_v6_protocol = {
1081         .handler        = dccp_v6_rcv,
1082         .err_handler    = dccp_v6_err,
1083         .flags          = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1084 };
1085
1086 static const struct proto_ops inet6_dccp_ops = {
1087         .family            = PF_INET6,
1088         .owner             = THIS_MODULE,
1089         .release           = inet6_release,
1090         .bind              = inet6_bind,
1091         .connect           = inet_stream_connect,
1092         .socketpair        = sock_no_socketpair,
1093         .accept            = inet_accept,
1094         .getname           = inet6_getname,
1095         .poll              = dccp_poll,
1096         .ioctl             = inet6_ioctl,
1097         .listen            = inet_dccp_listen,
1098         .shutdown          = inet_shutdown,
1099         .setsockopt        = sock_common_setsockopt,
1100         .getsockopt        = sock_common_getsockopt,
1101         .sendmsg           = inet_sendmsg,
1102         .recvmsg           = sock_common_recvmsg,
1103         .mmap              = sock_no_mmap,
1104         .sendpage          = sock_no_sendpage,
1105 #ifdef CONFIG_COMPAT
1106         .compat_setsockopt = compat_sock_common_setsockopt,
1107         .compat_getsockopt = compat_sock_common_getsockopt,
1108 #endif
1109 };
1110
1111 static struct inet_protosw dccp_v6_protosw = {
1112         .type           = SOCK_DCCP,
1113         .protocol       = IPPROTO_DCCP,
1114         .prot           = &dccp_v6_prot,
1115         .ops            = &inet6_dccp_ops,
1116         .flags          = INET_PROTOSW_ICSK,
1117 };
1118
1119 static int __net_init dccp_v6_init_net(struct net *net)
1120 {
1121         if (dccp_hashinfo.bhash == NULL)
1122                 return -ESOCKTNOSUPPORT;
1123
1124         return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1125                                     SOCK_DCCP, IPPROTO_DCCP, net);
1126 }
1127
1128 static void __net_exit dccp_v6_exit_net(struct net *net)
1129 {
1130         inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1131 }
1132
1133 static struct pernet_operations dccp_v6_ops = {
1134         .init   = dccp_v6_init_net,
1135         .exit   = dccp_v6_exit_net,
1136 };
1137
1138 static int __init dccp_v6_init(void)
1139 {
1140         int err = proto_register(&dccp_v6_prot, 1);
1141
1142         if (err != 0)
1143                 goto out;
1144
1145         err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1146         if (err != 0)
1147                 goto out_unregister_proto;
1148
1149         inet6_register_protosw(&dccp_v6_protosw);
1150
1151         err = register_pernet_subsys(&dccp_v6_ops);
1152         if (err != 0)
1153                 goto out_destroy_ctl_sock;
1154 out:
1155         return err;
1156
1157 out_destroy_ctl_sock:
1158         inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1159         inet6_unregister_protosw(&dccp_v6_protosw);
1160 out_unregister_proto:
1161         proto_unregister(&dccp_v6_prot);
1162         goto out;
1163 }
1164
1165 static void __exit dccp_v6_exit(void)
1166 {
1167         unregister_pernet_subsys(&dccp_v6_ops);
1168         inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1169         inet6_unregister_protosw(&dccp_v6_protosw);
1170         proto_unregister(&dccp_v6_prot);
1171 }
1172
1173 module_init(dccp_v6_init);
1174 module_exit(dccp_v6_exit);
1175
1176 /*
1177  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1178  * values directly, Also cover the case where the protocol is not specified,
1179  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1180  */
1181 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1182 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1183 MODULE_LICENSE("GPL");
1184 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1185 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");