GNU Linux-libre 4.19.286-gnu1
[releases.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 #include <linux/string.h>
20
21 #include <net/addrconf.h>
22 #include <net/inet_common.h>
23 #include <net/inet_hashtables.h>
24 #include <net/inet_sock.h>
25 #include <net/inet6_connection_sock.h>
26 #include <net/inet6_hashtables.h>
27 #include <net/ip6_route.h>
28 #include <net/ipv6.h>
29 #include <net/protocol.h>
30 #include <net/transp_v6.h>
31 #include <net/ip6_checksum.h>
32 #include <net/xfrm.h>
33 #include <net/secure_seq.h>
34 #include <net/sock.h>
35
36 #include "dccp.h"
37 #include "ipv6.h"
38 #include "feat.h"
39
40 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
41
42 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
43 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
44
45 /* add pseudo-header to DCCP checksum stored in skb->csum */
46 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
47                                       const struct in6_addr *saddr,
48                                       const struct in6_addr *daddr)
49 {
50         return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
51 }
52
53 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
54 {
55         struct ipv6_pinfo *np = inet6_sk(sk);
56         struct dccp_hdr *dh = dccp_hdr(skb);
57
58         dccp_csum_outgoing(skb);
59         dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
60 }
61
62 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
63 {
64         return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
65                                              ipv6_hdr(skb)->saddr.s6_addr32,
66                                              dccp_hdr(skb)->dccph_dport,
67                                              dccp_hdr(skb)->dccph_sport     );
68
69 }
70
71 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
72                         u8 type, u8 code, int offset, __be32 info)
73 {
74         const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
75         const struct dccp_hdr *dh;
76         struct dccp_sock *dp;
77         struct ipv6_pinfo *np;
78         struct sock *sk;
79         int err;
80         __u64 seq;
81         struct net *net = dev_net(skb->dev);
82
83         /* Only need dccph_dport & dccph_sport which are the first
84          * 4 bytes in dccp header.
85          * Our caller (icmpv6_notify()) already pulled 8 bytes for us.
86          */
87         BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8);
88         BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8);
89         dh = (struct dccp_hdr *)(skb->data + offset);
90
91         sk = __inet6_lookup_established(net, &dccp_hashinfo,
92                                         &hdr->daddr, dh->dccph_dport,
93                                         &hdr->saddr, ntohs(dh->dccph_sport),
94                                         inet6_iif(skb), 0);
95
96         if (!sk) {
97                 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
98                                   ICMP6_MIB_INERRORS);
99                 return;
100         }
101
102         if (sk->sk_state == DCCP_TIME_WAIT) {
103                 inet_twsk_put(inet_twsk(sk));
104                 return;
105         }
106         seq = dccp_hdr_seq(dh);
107         if (sk->sk_state == DCCP_NEW_SYN_RECV)
108                 return dccp_req_err(sk, seq);
109
110         bh_lock_sock(sk);
111         if (sock_owned_by_user(sk))
112                 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
113
114         if (sk->sk_state == DCCP_CLOSED)
115                 goto out;
116
117         dp = dccp_sk(sk);
118         if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
119             !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
120                 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
121                 goto out;
122         }
123
124         np = inet6_sk(sk);
125
126         if (type == NDISC_REDIRECT) {
127                 if (!sock_owned_by_user(sk)) {
128                         struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
129
130                         if (dst)
131                                 dst->ops->redirect(dst, sk, skb);
132                 }
133                 goto out;
134         }
135
136         if (type == ICMPV6_PKT_TOOBIG) {
137                 struct dst_entry *dst = NULL;
138
139                 if (!ip6_sk_accept_pmtu(sk))
140                         goto out;
141
142                 if (sock_owned_by_user(sk))
143                         goto out;
144                 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
145                         goto out;
146
147                 dst = inet6_csk_update_pmtu(sk, ntohl(info));
148                 if (!dst)
149                         goto out;
150
151                 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
152                         dccp_sync_mss(sk, dst_mtu(dst));
153                 goto out;
154         }
155
156         icmpv6_err_convert(type, code, &err);
157
158         /* Might be for an request_sock */
159         switch (sk->sk_state) {
160         case DCCP_REQUESTING:
161         case DCCP_RESPOND:  /* Cannot happen.
162                                It can, it SYNs are crossed. --ANK */
163                 if (!sock_owned_by_user(sk)) {
164                         __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
165                         sk->sk_err = err;
166                         /*
167                          * Wake people up to see the error
168                          * (see connect in sock.c)
169                          */
170                         sk->sk_error_report(sk);
171                         dccp_done(sk);
172                 } else
173                         sk->sk_err_soft = err;
174                 goto out;
175         }
176
177         if (!sock_owned_by_user(sk) && np->recverr) {
178                 sk->sk_err = err;
179                 sk->sk_error_report(sk);
180         } else
181                 sk->sk_err_soft = err;
182
183 out:
184         bh_unlock_sock(sk);
185         sock_put(sk);
186 }
187
188
189 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
190 {
191         struct inet_request_sock *ireq = inet_rsk(req);
192         struct ipv6_pinfo *np = inet6_sk(sk);
193         struct sk_buff *skb;
194         struct in6_addr *final_p, final;
195         struct flowi6 fl6;
196         int err = -1;
197         struct dst_entry *dst;
198
199         memset(&fl6, 0, sizeof(fl6));
200         fl6.flowi6_proto = IPPROTO_DCCP;
201         fl6.daddr = ireq->ir_v6_rmt_addr;
202         fl6.saddr = ireq->ir_v6_loc_addr;
203         fl6.flowlabel = 0;
204         fl6.flowi6_oif = ireq->ir_iif;
205         fl6.fl6_dport = ireq->ir_rmt_port;
206         fl6.fl6_sport = htons(ireq->ir_num);
207         security_req_classify_flow(req, flowi6_to_flowi(&fl6));
208
209
210         rcu_read_lock();
211         final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
212         rcu_read_unlock();
213
214         dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
215         if (IS_ERR(dst)) {
216                 err = PTR_ERR(dst);
217                 dst = NULL;
218                 goto done;
219         }
220
221         skb = dccp_make_response(sk, dst, req);
222         if (skb != NULL) {
223                 struct dccp_hdr *dh = dccp_hdr(skb);
224                 struct ipv6_txoptions *opt;
225
226                 dh->dccph_checksum = dccp_v6_csum_finish(skb,
227                                                          &ireq->ir_v6_loc_addr,
228                                                          &ireq->ir_v6_rmt_addr);
229                 fl6.daddr = ireq->ir_v6_rmt_addr;
230                 rcu_read_lock();
231                 opt = ireq->ipv6_opt;
232                 if (!opt)
233                         opt = rcu_dereference(np->opt);
234                 err = ip6_xmit(sk, skb, &fl6, sk->sk_mark, opt, np->tclass);
235                 rcu_read_unlock();
236                 err = net_xmit_eval(err);
237         }
238
239 done:
240         dst_release(dst);
241         return err;
242 }
243
244 static void dccp_v6_reqsk_destructor(struct request_sock *req)
245 {
246         dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
247         kfree(inet_rsk(req)->ipv6_opt);
248         kfree_skb(inet_rsk(req)->pktopts);
249 }
250
251 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
252 {
253         const struct ipv6hdr *rxip6h;
254         struct sk_buff *skb;
255         struct flowi6 fl6;
256         struct net *net = dev_net(skb_dst(rxskb)->dev);
257         struct sock *ctl_sk = net->dccp.v6_ctl_sk;
258         struct dst_entry *dst;
259
260         if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
261                 return;
262
263         if (!ipv6_unicast_destination(rxskb))
264                 return;
265
266         skb = dccp_ctl_make_reset(ctl_sk, rxskb);
267         if (skb == NULL)
268                 return;
269
270         rxip6h = ipv6_hdr(rxskb);
271         dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
272                                                             &rxip6h->daddr);
273
274         memset(&fl6, 0, sizeof(fl6));
275         fl6.daddr = rxip6h->saddr;
276         fl6.saddr = rxip6h->daddr;
277
278         fl6.flowi6_proto = IPPROTO_DCCP;
279         fl6.flowi6_oif = inet6_iif(rxskb);
280         fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
281         fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
282         security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
283
284         /* sk = NULL, but it is safe for now. RST socket required. */
285         dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL);
286         if (!IS_ERR(dst)) {
287                 skb_dst_set(skb, dst);
288                 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0);
289                 DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
290                 DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
291                 return;
292         }
293
294         kfree_skb(skb);
295 }
296
297 static struct request_sock_ops dccp6_request_sock_ops = {
298         .family         = AF_INET6,
299         .obj_size       = sizeof(struct dccp6_request_sock),
300         .rtx_syn_ack    = dccp_v6_send_response,
301         .send_ack       = dccp_reqsk_send_ack,
302         .destructor     = dccp_v6_reqsk_destructor,
303         .send_reset     = dccp_v6_ctl_send_reset,
304         .syn_ack_timeout = dccp_syn_ack_timeout,
305 };
306
307 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
308 {
309         struct request_sock *req;
310         struct dccp_request_sock *dreq;
311         struct inet_request_sock *ireq;
312         struct ipv6_pinfo *np = inet6_sk(sk);
313         const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
314         struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
315
316         if (skb->protocol == htons(ETH_P_IP))
317                 return dccp_v4_conn_request(sk, skb);
318
319         if (!ipv6_unicast_destination(skb))
320                 return 0;       /* discard, don't send a reset here */
321
322         if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) {
323                 __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS);
324                 return 0;
325         }
326
327         if (dccp_bad_service_code(sk, service)) {
328                 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
329                 goto drop;
330         }
331         /*
332          * There are no SYN attacks on IPv6, yet...
333          */
334         dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
335         if (inet_csk_reqsk_queue_is_full(sk))
336                 goto drop;
337
338         if (sk_acceptq_is_full(sk))
339                 goto drop;
340
341         req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
342         if (req == NULL)
343                 goto drop;
344
345         if (dccp_reqsk_init(req, dccp_sk(sk), skb))
346                 goto drop_and_free;
347
348         dreq = dccp_rsk(req);
349         if (dccp_parse_options(sk, dreq, skb))
350                 goto drop_and_free;
351
352         if (security_inet_conn_request(sk, skb, req))
353                 goto drop_and_free;
354
355         ireq = inet_rsk(req);
356         ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
357         ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
358         ireq->ireq_family = AF_INET6;
359         ireq->ir_mark = inet_request_mark(sk, skb);
360
361         if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
362             np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
363             np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
364                 refcount_inc(&skb->users);
365                 ireq->pktopts = skb;
366         }
367         ireq->ir_iif = sk->sk_bound_dev_if;
368
369         /* So that link locals have meaning */
370         if (!sk->sk_bound_dev_if &&
371             ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
372                 ireq->ir_iif = inet6_iif(skb);
373
374         /*
375          * Step 3: Process LISTEN state
376          *
377          *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
378          *
379          * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
380          */
381         dreq->dreq_isr     = dcb->dccpd_seq;
382         dreq->dreq_gsr     = dreq->dreq_isr;
383         dreq->dreq_iss     = dccp_v6_init_sequence(skb);
384         dreq->dreq_gss     = dreq->dreq_iss;
385         dreq->dreq_service = service;
386
387         if (dccp_v6_send_response(sk, req))
388                 goto drop_and_free;
389
390         inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
391         reqsk_put(req);
392         return 0;
393
394 drop_and_free:
395         reqsk_free(req);
396 drop:
397         __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
398         return -1;
399 }
400
401 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
402                                               struct sk_buff *skb,
403                                               struct request_sock *req,
404                                               struct dst_entry *dst,
405                                               struct request_sock *req_unhash,
406                                               bool *own_req)
407 {
408         struct inet_request_sock *ireq = inet_rsk(req);
409         struct ipv6_pinfo *newnp;
410         const struct ipv6_pinfo *np = inet6_sk(sk);
411         struct ipv6_txoptions *opt;
412         struct inet_sock *newinet;
413         struct dccp6_sock *newdp6;
414         struct sock *newsk;
415
416         if (skb->protocol == htons(ETH_P_IP)) {
417                 /*
418                  *      v6 mapped
419                  */
420                 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
421                                                   req_unhash, own_req);
422                 if (newsk == NULL)
423                         return NULL;
424
425                 newdp6 = (struct dccp6_sock *)newsk;
426                 newinet = inet_sk(newsk);
427                 newinet->pinet6 = &newdp6->inet6;
428                 newnp = inet6_sk(newsk);
429
430                 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
431
432                 newnp->saddr = newsk->sk_v6_rcv_saddr;
433
434                 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
435                 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
436                 newnp->pktoptions  = NULL;
437                 newnp->opt         = NULL;
438                 newnp->ipv6_mc_list = NULL;
439                 newnp->ipv6_ac_list = NULL;
440                 newnp->ipv6_fl_list = NULL;
441                 newnp->mcast_oif   = inet_iif(skb);
442                 newnp->mcast_hops  = ip_hdr(skb)->ttl;
443
444                 /*
445                  * No need to charge this sock to the relevant IPv6 refcnt debug socks count
446                  * here, dccp_create_openreq_child now does this for us, see the comment in
447                  * that function for the gory details. -acme
448                  */
449
450                 /* It is tricky place. Until this moment IPv4 tcp
451                    worked with IPv6 icsk.icsk_af_ops.
452                    Sync it now.
453                  */
454                 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
455
456                 return newsk;
457         }
458
459
460         if (sk_acceptq_is_full(sk))
461                 goto out_overflow;
462
463         if (!dst) {
464                 struct flowi6 fl6;
465
466                 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
467                 if (!dst)
468                         goto out;
469         }
470
471         newsk = dccp_create_openreq_child(sk, req, skb);
472         if (newsk == NULL)
473                 goto out_nonewsk;
474
475         /*
476          * No need to charge this sock to the relevant IPv6 refcnt debug socks
477          * count here, dccp_create_openreq_child now does this for us, see the
478          * comment in that function for the gory details. -acme
479          */
480
481         ip6_dst_store(newsk, dst, NULL, NULL);
482         newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
483                                                       NETIF_F_TSO);
484         newdp6 = (struct dccp6_sock *)newsk;
485         newinet = inet_sk(newsk);
486         newinet->pinet6 = &newdp6->inet6;
487         newnp = inet6_sk(newsk);
488
489         memcpy(newnp, np, sizeof(struct ipv6_pinfo));
490
491         newsk->sk_v6_daddr      = ireq->ir_v6_rmt_addr;
492         newnp->saddr            = ireq->ir_v6_loc_addr;
493         newsk->sk_v6_rcv_saddr  = ireq->ir_v6_loc_addr;
494         newsk->sk_bound_dev_if  = ireq->ir_iif;
495
496         /* Now IPv6 options...
497
498            First: no IPv4 options.
499          */
500         newinet->inet_opt = NULL;
501
502         /* Clone RX bits */
503         newnp->rxopt.all = np->rxopt.all;
504
505         newnp->ipv6_mc_list = NULL;
506         newnp->ipv6_ac_list = NULL;
507         newnp->ipv6_fl_list = NULL;
508         newnp->pktoptions = NULL;
509         newnp->opt        = NULL;
510         newnp->mcast_oif  = inet6_iif(skb);
511         newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
512
513         /*
514          * Clone native IPv6 options from listening socket (if any)
515          *
516          * Yes, keeping reference count would be much more clever, but we make
517          * one more one thing there: reattach optmem to newsk.
518          */
519         opt = ireq->ipv6_opt;
520         if (!opt)
521                 opt = rcu_dereference(np->opt);
522         if (opt) {
523                 opt = ipv6_dup_options(newsk, opt);
524                 RCU_INIT_POINTER(newnp->opt, opt);
525         }
526         inet_csk(newsk)->icsk_ext_hdr_len = 0;
527         if (opt)
528                 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
529                                                     opt->opt_flen;
530
531         dccp_sync_mss(newsk, dst_mtu(dst));
532
533         newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
534         newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
535
536         if (__inet_inherit_port(sk, newsk) < 0) {
537                 inet_csk_prepare_forced_close(newsk);
538                 dccp_done(newsk);
539                 goto out;
540         }
541         *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
542         /* Clone pktoptions received with SYN, if we own the req */
543         if (*own_req && ireq->pktopts) {
544                 newnp->pktoptions = skb_clone_and_charge_r(ireq->pktopts, newsk);
545                 consume_skb(ireq->pktopts);
546                 ireq->pktopts = NULL;
547         }
548
549         return newsk;
550
551 out_overflow:
552         __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
553 out_nonewsk:
554         dst_release(dst);
555 out:
556         __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
557         return NULL;
558 }
559
560 /* The socket must have it's spinlock held when we get
561  * here.
562  *
563  * We have a potential double-lock case here, so even when
564  * doing backlog processing we use the BH locking scheme.
565  * This is because we cannot sleep with the original spinlock
566  * held.
567  */
568 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
569 {
570         struct ipv6_pinfo *np = inet6_sk(sk);
571         struct sk_buff *opt_skb = NULL;
572
573         /* Imagine: socket is IPv6. IPv4 packet arrives,
574            goes to IPv4 receive handler and backlogged.
575            From backlog it always goes here. Kerboom...
576            Fortunately, dccp_rcv_established and rcv_established
577            handle them correctly, but it is not case with
578            dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
579          */
580
581         if (skb->protocol == htons(ETH_P_IP))
582                 return dccp_v4_do_rcv(sk, skb);
583
584         if (sk_filter(sk, skb))
585                 goto discard;
586
587         /*
588          * socket locking is here for SMP purposes as backlog rcv is currently
589          * called with bh processing disabled.
590          */
591
592         /* Do Stevens' IPV6_PKTOPTIONS.
593
594            Yes, guys, it is the only place in our code, where we
595            may make it not affecting IPv4.
596            The rest of code is protocol independent,
597            and I do not like idea to uglify IPv4.
598
599            Actually, all the idea behind IPV6_PKTOPTIONS
600            looks not very well thought. For now we latch
601            options, received in the last packet, enqueued
602            by tcp. Feel free to propose better solution.
603                                                --ANK (980728)
604          */
605         if (np->rxopt.all)
606                 opt_skb = skb_clone_and_charge_r(skb, sk);
607
608         if (sk->sk_state == DCCP_OPEN) { /* Fast path */
609                 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
610                         goto reset;
611                 if (opt_skb)
612                         goto ipv6_pktoptions;
613                 return 0;
614         }
615
616         /*
617          *  Step 3: Process LISTEN state
618          *     If S.state == LISTEN,
619          *       If P.type == Request or P contains a valid Init Cookie option,
620          *            (* Must scan the packet's options to check for Init
621          *               Cookies.  Only Init Cookies are processed here,
622          *               however; other options are processed in Step 8.  This
623          *               scan need only be performed if the endpoint uses Init
624          *               Cookies *)
625          *            (* Generate a new socket and switch to that socket *)
626          *            Set S := new socket for this port pair
627          *            S.state = RESPOND
628          *            Choose S.ISS (initial seqno) or set from Init Cookies
629          *            Initialize S.GAR := S.ISS
630          *            Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
631          *            Continue with S.state == RESPOND
632          *            (* A Response packet will be generated in Step 11 *)
633          *       Otherwise,
634          *            Generate Reset(No Connection) unless P.type == Reset
635          *            Drop packet and return
636          *
637          * NOTE: the check for the packet types is done in
638          *       dccp_rcv_state_process
639          */
640
641         if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
642                 goto reset;
643         if (opt_skb)
644                 goto ipv6_pktoptions;
645         return 0;
646
647 reset:
648         dccp_v6_ctl_send_reset(sk, skb);
649 discard:
650         if (opt_skb != NULL)
651                 __kfree_skb(opt_skb);
652         kfree_skb(skb);
653         return 0;
654
655 /* Handling IPV6_PKTOPTIONS skb the similar
656  * way it's done for net/ipv6/tcp_ipv6.c
657  */
658 ipv6_pktoptions:
659         if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
660                 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
661                         np->mcast_oif = inet6_iif(opt_skb);
662                 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
663                         np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
664                 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
665                         np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
666                 if (np->repflow)
667                         np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
668                 if (ipv6_opt_accepted(sk, opt_skb,
669                                       &DCCP_SKB_CB(opt_skb)->header.h6)) {
670                         memmove(IP6CB(opt_skb),
671                                 &DCCP_SKB_CB(opt_skb)->header.h6,
672                                 sizeof(struct inet6_skb_parm));
673                         opt_skb = xchg(&np->pktoptions, opt_skb);
674                 } else {
675                         __kfree_skb(opt_skb);
676                         opt_skb = xchg(&np->pktoptions, NULL);
677                 }
678         }
679
680         kfree_skb(opt_skb);
681         return 0;
682 }
683
684 static int dccp_v6_rcv(struct sk_buff *skb)
685 {
686         const struct dccp_hdr *dh;
687         bool refcounted;
688         struct sock *sk;
689         int min_cov;
690
691         /* Step 1: Check header basics */
692
693         if (dccp_invalid_packet(skb))
694                 goto discard_it;
695
696         /* Step 1: If header checksum is incorrect, drop packet and return. */
697         if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
698                                      &ipv6_hdr(skb)->daddr)) {
699                 DCCP_WARN("dropped packet with invalid checksum\n");
700                 goto discard_it;
701         }
702
703         dh = dccp_hdr(skb);
704
705         DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
706         DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
707
708         if (dccp_packet_without_ack(skb))
709                 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
710         else
711                 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
712
713 lookup:
714         sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
715                                 dh->dccph_sport, dh->dccph_dport,
716                                 inet6_iif(skb), 0, &refcounted);
717         if (!sk) {
718                 dccp_pr_debug("failed to look up flow ID in table and "
719                               "get corresponding socket\n");
720                 goto no_dccp_socket;
721         }
722
723         /*
724          * Step 2:
725          *      ... or S.state == TIMEWAIT,
726          *              Generate Reset(No Connection) unless P.type == Reset
727          *              Drop packet and return
728          */
729         if (sk->sk_state == DCCP_TIME_WAIT) {
730                 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
731                 inet_twsk_put(inet_twsk(sk));
732                 goto no_dccp_socket;
733         }
734
735         if (sk->sk_state == DCCP_NEW_SYN_RECV) {
736                 struct request_sock *req = inet_reqsk(sk);
737                 struct sock *nsk;
738
739                 sk = req->rsk_listener;
740                 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
741                         inet_csk_reqsk_queue_drop_and_put(sk, req);
742                         goto lookup;
743                 }
744                 sock_hold(sk);
745                 refcounted = true;
746                 nsk = dccp_check_req(sk, skb, req);
747                 if (!nsk) {
748                         reqsk_put(req);
749                         goto discard_and_relse;
750                 }
751                 if (nsk == sk) {
752                         reqsk_put(req);
753                 } else if (dccp_child_process(sk, nsk, skb)) {
754                         dccp_v6_ctl_send_reset(sk, skb);
755                         goto discard_and_relse;
756                 } else {
757                         sock_put(sk);
758                         return 0;
759                 }
760         }
761         /*
762          * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
763          *      o if MinCsCov = 0, only packets with CsCov = 0 are accepted
764          *      o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
765          */
766         min_cov = dccp_sk(sk)->dccps_pcrlen;
767         if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
768                 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
769                               dh->dccph_cscov, min_cov);
770                 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
771                 goto discard_and_relse;
772         }
773
774         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
775                 goto discard_and_relse;
776
777         return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
778                                 refcounted) ? -1 : 0;
779
780 no_dccp_socket:
781         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
782                 goto discard_it;
783         /*
784          * Step 2:
785          *      If no socket ...
786          *              Generate Reset(No Connection) unless P.type == Reset
787          *              Drop packet and return
788          */
789         if (dh->dccph_type != DCCP_PKT_RESET) {
790                 DCCP_SKB_CB(skb)->dccpd_reset_code =
791                                         DCCP_RESET_CODE_NO_CONNECTION;
792                 dccp_v6_ctl_send_reset(sk, skb);
793         }
794
795 discard_it:
796         kfree_skb(skb);
797         return 0;
798
799 discard_and_relse:
800         if (refcounted)
801                 sock_put(sk);
802         goto discard_it;
803 }
804
805 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
806                            int addr_len)
807 {
808         struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
809         struct inet_connection_sock *icsk = inet_csk(sk);
810         struct inet_sock *inet = inet_sk(sk);
811         struct ipv6_pinfo *np = inet6_sk(sk);
812         struct dccp_sock *dp = dccp_sk(sk);
813         struct in6_addr *saddr = NULL, *final_p, final;
814         struct ipv6_txoptions *opt;
815         struct flowi6 fl6;
816         struct dst_entry *dst;
817         int addr_type;
818         int err;
819
820         dp->dccps_role = DCCP_ROLE_CLIENT;
821
822         if (addr_len < SIN6_LEN_RFC2133)
823                 return -EINVAL;
824
825         if (usin->sin6_family != AF_INET6)
826                 return -EAFNOSUPPORT;
827
828         memset(&fl6, 0, sizeof(fl6));
829
830         if (np->sndflow) {
831                 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
832                 IP6_ECN_flow_init(fl6.flowlabel);
833                 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
834                         struct ip6_flowlabel *flowlabel;
835                         flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
836                         if (flowlabel == NULL)
837                                 return -EINVAL;
838                         fl6_sock_release(flowlabel);
839                 }
840         }
841         /*
842          * connect() to INADDR_ANY means loopback (BSD'ism).
843          */
844         if (ipv6_addr_any(&usin->sin6_addr))
845                 usin->sin6_addr.s6_addr[15] = 1;
846
847         addr_type = ipv6_addr_type(&usin->sin6_addr);
848
849         if (addr_type & IPV6_ADDR_MULTICAST)
850                 return -ENETUNREACH;
851
852         if (addr_type & IPV6_ADDR_LINKLOCAL) {
853                 if (addr_len >= sizeof(struct sockaddr_in6) &&
854                     usin->sin6_scope_id) {
855                         /* If interface is set while binding, indices
856                          * must coincide.
857                          */
858                         if (sk->sk_bound_dev_if &&
859                             sk->sk_bound_dev_if != usin->sin6_scope_id)
860                                 return -EINVAL;
861
862                         sk->sk_bound_dev_if = usin->sin6_scope_id;
863                 }
864
865                 /* Connect to link-local address requires an interface */
866                 if (!sk->sk_bound_dev_if)
867                         return -EINVAL;
868         }
869
870         sk->sk_v6_daddr = usin->sin6_addr;
871         np->flow_label = fl6.flowlabel;
872
873         /*
874          * DCCP over IPv4
875          */
876         if (addr_type == IPV6_ADDR_MAPPED) {
877                 u32 exthdrlen = icsk->icsk_ext_hdr_len;
878                 struct sockaddr_in sin;
879
880                 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
881
882                 if (__ipv6_only_sock(sk))
883                         return -ENETUNREACH;
884
885                 sin.sin_family = AF_INET;
886                 sin.sin_port = usin->sin6_port;
887                 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
888
889                 icsk->icsk_af_ops = &dccp_ipv6_mapped;
890                 sk->sk_backlog_rcv = dccp_v4_do_rcv;
891
892                 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
893                 if (err) {
894                         icsk->icsk_ext_hdr_len = exthdrlen;
895                         icsk->icsk_af_ops = &dccp_ipv6_af_ops;
896                         sk->sk_backlog_rcv = dccp_v6_do_rcv;
897                         goto failure;
898                 }
899                 np->saddr = sk->sk_v6_rcv_saddr;
900                 return err;
901         }
902
903         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
904                 saddr = &sk->sk_v6_rcv_saddr;
905
906         fl6.flowi6_proto = IPPROTO_DCCP;
907         fl6.daddr = sk->sk_v6_daddr;
908         fl6.saddr = saddr ? *saddr : np->saddr;
909         fl6.flowi6_oif = sk->sk_bound_dev_if;
910         fl6.fl6_dport = usin->sin6_port;
911         fl6.fl6_sport = inet->inet_sport;
912         security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
913
914         opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
915         final_p = fl6_update_dst(&fl6, opt, &final);
916
917         dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
918         if (IS_ERR(dst)) {
919                 err = PTR_ERR(dst);
920                 goto failure;
921         }
922
923         if (saddr == NULL) {
924                 saddr = &fl6.saddr;
925                 sk->sk_v6_rcv_saddr = *saddr;
926         }
927
928         /* set the source address */
929         np->saddr = *saddr;
930         inet->inet_rcv_saddr = LOOPBACK4_IPV6;
931
932         ip6_dst_store(sk, dst, NULL, NULL);
933
934         icsk->icsk_ext_hdr_len = 0;
935         if (opt)
936                 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
937
938         inet->inet_dport = usin->sin6_port;
939
940         dccp_set_state(sk, DCCP_REQUESTING);
941         err = inet6_hash_connect(&dccp_death_row, sk);
942         if (err)
943                 goto late_failure;
944
945         dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
946                                                       sk->sk_v6_daddr.s6_addr32,
947                                                       inet->inet_sport,
948                                                       inet->inet_dport);
949         err = dccp_connect(sk);
950         if (err)
951                 goto late_failure;
952
953         return 0;
954
955 late_failure:
956         dccp_set_state(sk, DCCP_CLOSED);
957         if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
958                 inet_reset_saddr(sk);
959         __sk_dst_reset(sk);
960 failure:
961         inet->inet_dport = 0;
962         sk->sk_route_caps = 0;
963         return err;
964 }
965
966 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
967         .queue_xmit        = inet6_csk_xmit,
968         .send_check        = dccp_v6_send_check,
969         .rebuild_header    = inet6_sk_rebuild_header,
970         .conn_request      = dccp_v6_conn_request,
971         .syn_recv_sock     = dccp_v6_request_recv_sock,
972         .net_header_len    = sizeof(struct ipv6hdr),
973         .setsockopt        = ipv6_setsockopt,
974         .getsockopt        = ipv6_getsockopt,
975         .addr2sockaddr     = inet6_csk_addr2sockaddr,
976         .sockaddr_len      = sizeof(struct sockaddr_in6),
977 #ifdef CONFIG_COMPAT
978         .compat_setsockopt = compat_ipv6_setsockopt,
979         .compat_getsockopt = compat_ipv6_getsockopt,
980 #endif
981 };
982
983 /*
984  *      DCCP over IPv4 via INET6 API
985  */
986 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
987         .queue_xmit        = ip_queue_xmit,
988         .send_check        = dccp_v4_send_check,
989         .rebuild_header    = inet_sk_rebuild_header,
990         .conn_request      = dccp_v6_conn_request,
991         .syn_recv_sock     = dccp_v6_request_recv_sock,
992         .net_header_len    = sizeof(struct iphdr),
993         .setsockopt        = ipv6_setsockopt,
994         .getsockopt        = ipv6_getsockopt,
995         .addr2sockaddr     = inet6_csk_addr2sockaddr,
996         .sockaddr_len      = sizeof(struct sockaddr_in6),
997 #ifdef CONFIG_COMPAT
998         .compat_setsockopt = compat_ipv6_setsockopt,
999         .compat_getsockopt = compat_ipv6_getsockopt,
1000 #endif
1001 };
1002
1003 static void dccp_v6_sk_destruct(struct sock *sk)
1004 {
1005         dccp_destruct_common(sk);
1006         inet6_sock_destruct(sk);
1007 }
1008
1009 /* NOTE: A lot of things set to zero explicitly by call to
1010  *       sk_alloc() so need not be done here.
1011  */
1012 static int dccp_v6_init_sock(struct sock *sk)
1013 {
1014         static __u8 dccp_v6_ctl_sock_initialized;
1015         int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1016
1017         if (err == 0) {
1018                 if (unlikely(!dccp_v6_ctl_sock_initialized))
1019                         dccp_v6_ctl_sock_initialized = 1;
1020                 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1021                 sk->sk_destruct = dccp_v6_sk_destruct;
1022         }
1023
1024         return err;
1025 }
1026
1027 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1028         .twsk_obj_size  = sizeof(struct dccp6_timewait_sock),
1029 };
1030
1031 static struct proto dccp_v6_prot = {
1032         .name              = "DCCPv6",
1033         .owner             = THIS_MODULE,
1034         .close             = dccp_close,
1035         .connect           = dccp_v6_connect,
1036         .disconnect        = dccp_disconnect,
1037         .ioctl             = dccp_ioctl,
1038         .init              = dccp_v6_init_sock,
1039         .setsockopt        = dccp_setsockopt,
1040         .getsockopt        = dccp_getsockopt,
1041         .sendmsg           = dccp_sendmsg,
1042         .recvmsg           = dccp_recvmsg,
1043         .backlog_rcv       = dccp_v6_do_rcv,
1044         .hash              = inet6_hash,
1045         .unhash            = inet_unhash,
1046         .accept            = inet_csk_accept,
1047         .get_port          = inet_csk_get_port,
1048         .shutdown          = dccp_shutdown,
1049         .destroy           = dccp_destroy_sock,
1050         .orphan_count      = &dccp_orphan_count,
1051         .max_header        = MAX_DCCP_HEADER,
1052         .obj_size          = sizeof(struct dccp6_sock),
1053         .slab_flags        = SLAB_TYPESAFE_BY_RCU,
1054         .rsk_prot          = &dccp6_request_sock_ops,
1055         .twsk_prot         = &dccp6_timewait_sock_ops,
1056         .h.hashinfo        = &dccp_hashinfo,
1057 #ifdef CONFIG_COMPAT
1058         .compat_setsockopt = compat_dccp_setsockopt,
1059         .compat_getsockopt = compat_dccp_getsockopt,
1060 #endif
1061 };
1062
1063 static const struct inet6_protocol dccp_v6_protocol = {
1064         .handler        = dccp_v6_rcv,
1065         .err_handler    = dccp_v6_err,
1066         .flags          = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1067 };
1068
1069 static const struct proto_ops inet6_dccp_ops = {
1070         .family            = PF_INET6,
1071         .owner             = THIS_MODULE,
1072         .release           = inet6_release,
1073         .bind              = inet6_bind,
1074         .connect           = inet_stream_connect,
1075         .socketpair        = sock_no_socketpair,
1076         .accept            = inet_accept,
1077         .getname           = inet6_getname,
1078         .poll              = dccp_poll,
1079         .ioctl             = inet6_ioctl,
1080         .listen            = inet_dccp_listen,
1081         .shutdown          = inet_shutdown,
1082         .setsockopt        = sock_common_setsockopt,
1083         .getsockopt        = sock_common_getsockopt,
1084         .sendmsg           = inet_sendmsg,
1085         .recvmsg           = sock_common_recvmsg,
1086         .mmap              = sock_no_mmap,
1087         .sendpage          = sock_no_sendpage,
1088 #ifdef CONFIG_COMPAT
1089         .compat_setsockopt = compat_sock_common_setsockopt,
1090         .compat_getsockopt = compat_sock_common_getsockopt,
1091 #endif
1092 };
1093
1094 static struct inet_protosw dccp_v6_protosw = {
1095         .type           = SOCK_DCCP,
1096         .protocol       = IPPROTO_DCCP,
1097         .prot           = &dccp_v6_prot,
1098         .ops            = &inet6_dccp_ops,
1099         .flags          = INET_PROTOSW_ICSK,
1100 };
1101
1102 static int __net_init dccp_v6_init_net(struct net *net)
1103 {
1104         if (dccp_hashinfo.bhash == NULL)
1105                 return -ESOCKTNOSUPPORT;
1106
1107         return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1108                                     SOCK_DCCP, IPPROTO_DCCP, net);
1109 }
1110
1111 static void __net_exit dccp_v6_exit_net(struct net *net)
1112 {
1113         inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1114 }
1115
1116 static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list)
1117 {
1118         inet_twsk_purge(&dccp_hashinfo, AF_INET6);
1119 }
1120
1121 static struct pernet_operations dccp_v6_ops = {
1122         .init   = dccp_v6_init_net,
1123         .exit   = dccp_v6_exit_net,
1124         .exit_batch = dccp_v6_exit_batch,
1125 };
1126
1127 static int __init dccp_v6_init(void)
1128 {
1129         int err = proto_register(&dccp_v6_prot, 1);
1130
1131         if (err)
1132                 goto out;
1133
1134         inet6_register_protosw(&dccp_v6_protosw);
1135
1136         err = register_pernet_subsys(&dccp_v6_ops);
1137         if (err)
1138                 goto out_destroy_ctl_sock;
1139
1140         err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1141         if (err)
1142                 goto out_unregister_proto;
1143
1144 out:
1145         return err;
1146 out_unregister_proto:
1147         unregister_pernet_subsys(&dccp_v6_ops);
1148 out_destroy_ctl_sock:
1149         inet6_unregister_protosw(&dccp_v6_protosw);
1150         proto_unregister(&dccp_v6_prot);
1151         goto out;
1152 }
1153
1154 static void __exit dccp_v6_exit(void)
1155 {
1156         inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1157         unregister_pernet_subsys(&dccp_v6_ops);
1158         inet6_unregister_protosw(&dccp_v6_protosw);
1159         proto_unregister(&dccp_v6_prot);
1160 }
1161
1162 module_init(dccp_v6_init);
1163 module_exit(dccp_v6_exit);
1164
1165 /*
1166  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1167  * values directly, Also cover the case where the protocol is not specified,
1168  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1169  */
1170 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1171 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1172 MODULE_LICENSE("GPL");
1173 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1174 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");