Merge tag 'acpi-4.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
[sfrench/cifs-2.6.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(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(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 (dccp_bad_service_code(sk, service)) {
323                 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
324                 goto drop;
325         }
326         /*
327          * There are no SYN attacks on IPv6, yet...
328          */
329         dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
330         if (inet_csk_reqsk_queue_is_full(sk))
331                 goto drop;
332
333         if (sk_acceptq_is_full(sk))
334                 goto drop;
335
336         req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
337         if (req == NULL)
338                 goto drop;
339
340         if (dccp_reqsk_init(req, dccp_sk(sk), skb))
341                 goto drop_and_free;
342
343         dreq = dccp_rsk(req);
344         if (dccp_parse_options(sk, dreq, skb))
345                 goto drop_and_free;
346
347         if (security_inet_conn_request(sk, skb, req))
348                 goto drop_and_free;
349
350         ireq = inet_rsk(req);
351         ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
352         ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
353         ireq->ireq_family = AF_INET6;
354
355         if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
356             np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
357             np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
358                 refcount_inc(&skb->users);
359                 ireq->pktopts = skb;
360         }
361         ireq->ir_iif = sk->sk_bound_dev_if;
362
363         /* So that link locals have meaning */
364         if (!sk->sk_bound_dev_if &&
365             ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
366                 ireq->ir_iif = inet6_iif(skb);
367
368         /*
369          * Step 3: Process LISTEN state
370          *
371          *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
372          *
373          * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
374          */
375         dreq->dreq_isr     = dcb->dccpd_seq;
376         dreq->dreq_gsr     = dreq->dreq_isr;
377         dreq->dreq_iss     = dccp_v6_init_sequence(skb);
378         dreq->dreq_gss     = dreq->dreq_iss;
379         dreq->dreq_service = service;
380
381         if (dccp_v6_send_response(sk, req))
382                 goto drop_and_free;
383
384         inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
385         reqsk_put(req);
386         return 0;
387
388 drop_and_free:
389         reqsk_free(req);
390 drop:
391         __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
392         return -1;
393 }
394
395 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
396                                               struct sk_buff *skb,
397                                               struct request_sock *req,
398                                               struct dst_entry *dst,
399                                               struct request_sock *req_unhash,
400                                               bool *own_req)
401 {
402         struct inet_request_sock *ireq = inet_rsk(req);
403         struct ipv6_pinfo *newnp;
404         const struct ipv6_pinfo *np = inet6_sk(sk);
405         struct ipv6_txoptions *opt;
406         struct inet_sock *newinet;
407         struct dccp6_sock *newdp6;
408         struct sock *newsk;
409
410         if (skb->protocol == htons(ETH_P_IP)) {
411                 /*
412                  *      v6 mapped
413                  */
414                 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
415                                                   req_unhash, own_req);
416                 if (newsk == NULL)
417                         return NULL;
418
419                 newdp6 = (struct dccp6_sock *)newsk;
420                 newinet = inet_sk(newsk);
421                 newinet->pinet6 = &newdp6->inet6;
422                 newnp = inet6_sk(newsk);
423
424                 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
425
426                 newnp->saddr = newsk->sk_v6_rcv_saddr;
427
428                 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
429                 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
430                 newnp->pktoptions  = NULL;
431                 newnp->opt         = NULL;
432                 newnp->ipv6_mc_list = NULL;
433                 newnp->ipv6_ac_list = NULL;
434                 newnp->ipv6_fl_list = NULL;
435                 newnp->mcast_oif   = inet6_iif(skb);
436                 newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
437
438                 /*
439                  * No need to charge this sock to the relevant IPv6 refcnt debug socks count
440                  * here, dccp_create_openreq_child now does this for us, see the comment in
441                  * that function for the gory details. -acme
442                  */
443
444                 /* It is tricky place. Until this moment IPv4 tcp
445                    worked with IPv6 icsk.icsk_af_ops.
446                    Sync it now.
447                  */
448                 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
449
450                 return newsk;
451         }
452
453
454         if (sk_acceptq_is_full(sk))
455                 goto out_overflow;
456
457         if (!dst) {
458                 struct flowi6 fl6;
459
460                 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
461                 if (!dst)
462                         goto out;
463         }
464
465         newsk = dccp_create_openreq_child(sk, req, skb);
466         if (newsk == NULL)
467                 goto out_nonewsk;
468
469         /*
470          * No need to charge this sock to the relevant IPv6 refcnt debug socks
471          * count here, dccp_create_openreq_child now does this for us, see the
472          * comment in that function for the gory details. -acme
473          */
474
475         ip6_dst_store(newsk, dst, NULL, NULL);
476         newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
477                                                       NETIF_F_TSO);
478         newdp6 = (struct dccp6_sock *)newsk;
479         newinet = inet_sk(newsk);
480         newinet->pinet6 = &newdp6->inet6;
481         newnp = inet6_sk(newsk);
482
483         memcpy(newnp, np, sizeof(struct ipv6_pinfo));
484
485         newsk->sk_v6_daddr      = ireq->ir_v6_rmt_addr;
486         newnp->saddr            = ireq->ir_v6_loc_addr;
487         newsk->sk_v6_rcv_saddr  = ireq->ir_v6_loc_addr;
488         newsk->sk_bound_dev_if  = ireq->ir_iif;
489
490         /* Now IPv6 options...
491
492            First: no IPv4 options.
493          */
494         newinet->inet_opt = NULL;
495
496         /* Clone RX bits */
497         newnp->rxopt.all = np->rxopt.all;
498
499         newnp->ipv6_mc_list = NULL;
500         newnp->ipv6_ac_list = NULL;
501         newnp->ipv6_fl_list = NULL;
502         newnp->pktoptions = NULL;
503         newnp->opt        = NULL;
504         newnp->mcast_oif  = inet6_iif(skb);
505         newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
506
507         /*
508          * Clone native IPv6 options from listening socket (if any)
509          *
510          * Yes, keeping reference count would be much more clever, but we make
511          * one more one thing there: reattach optmem to newsk.
512          */
513         opt = ireq->ipv6_opt;
514         if (!opt)
515                 opt = rcu_dereference(np->opt);
516         if (opt) {
517                 opt = ipv6_dup_options(newsk, opt);
518                 RCU_INIT_POINTER(newnp->opt, opt);
519         }
520         inet_csk(newsk)->icsk_ext_hdr_len = 0;
521         if (opt)
522                 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
523                                                     opt->opt_flen;
524
525         dccp_sync_mss(newsk, dst_mtu(dst));
526
527         newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
528         newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
529
530         if (__inet_inherit_port(sk, newsk) < 0) {
531                 inet_csk_prepare_forced_close(newsk);
532                 dccp_done(newsk);
533                 goto out;
534         }
535         *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
536         /* Clone pktoptions received with SYN, if we own the req */
537         if (*own_req && ireq->pktopts) {
538                 newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
539                 consume_skb(ireq->pktopts);
540                 ireq->pktopts = NULL;
541                 if (newnp->pktoptions)
542                         skb_set_owner_r(newnp->pktoptions, newsk);
543         }
544
545         return newsk;
546
547 out_overflow:
548         __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
549 out_nonewsk:
550         dst_release(dst);
551 out:
552         __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
553         return NULL;
554 }
555
556 /* The socket must have it's spinlock held when we get
557  * here.
558  *
559  * We have a potential double-lock case here, so even when
560  * doing backlog processing we use the BH locking scheme.
561  * This is because we cannot sleep with the original spinlock
562  * held.
563  */
564 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
565 {
566         struct ipv6_pinfo *np = inet6_sk(sk);
567         struct sk_buff *opt_skb = NULL;
568
569         /* Imagine: socket is IPv6. IPv4 packet arrives,
570            goes to IPv4 receive handler and backlogged.
571            From backlog it always goes here. Kerboom...
572            Fortunately, dccp_rcv_established and rcv_established
573            handle them correctly, but it is not case with
574            dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
575          */
576
577         if (skb->protocol == htons(ETH_P_IP))
578                 return dccp_v4_do_rcv(sk, skb);
579
580         if (sk_filter(sk, skb))
581                 goto discard;
582
583         /*
584          * socket locking is here for SMP purposes as backlog rcv is currently
585          * called with bh processing disabled.
586          */
587
588         /* Do Stevens' IPV6_PKTOPTIONS.
589
590            Yes, guys, it is the only place in our code, where we
591            may make it not affecting IPv4.
592            The rest of code is protocol independent,
593            and I do not like idea to uglify IPv4.
594
595            Actually, all the idea behind IPV6_PKTOPTIONS
596            looks not very well thought. For now we latch
597            options, received in the last packet, enqueued
598            by tcp. Feel free to propose better solution.
599                                                --ANK (980728)
600          */
601         if (np->rxopt.all)
602                 opt_skb = skb_clone(skb, GFP_ATOMIC);
603
604         if (sk->sk_state == DCCP_OPEN) { /* Fast path */
605                 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
606                         goto reset;
607                 if (opt_skb)
608                         goto ipv6_pktoptions;
609                 return 0;
610         }
611
612         /*
613          *  Step 3: Process LISTEN state
614          *     If S.state == LISTEN,
615          *       If P.type == Request or P contains a valid Init Cookie option,
616          *            (* Must scan the packet's options to check for Init
617          *               Cookies.  Only Init Cookies are processed here,
618          *               however; other options are processed in Step 8.  This
619          *               scan need only be performed if the endpoint uses Init
620          *               Cookies *)
621          *            (* Generate a new socket and switch to that socket *)
622          *            Set S := new socket for this port pair
623          *            S.state = RESPOND
624          *            Choose S.ISS (initial seqno) or set from Init Cookies
625          *            Initialize S.GAR := S.ISS
626          *            Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
627          *            Continue with S.state == RESPOND
628          *            (* A Response packet will be generated in Step 11 *)
629          *       Otherwise,
630          *            Generate Reset(No Connection) unless P.type == Reset
631          *            Drop packet and return
632          *
633          * NOTE: the check for the packet types is done in
634          *       dccp_rcv_state_process
635          */
636
637         if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
638                 goto reset;
639         if (opt_skb)
640                 goto ipv6_pktoptions;
641         return 0;
642
643 reset:
644         dccp_v6_ctl_send_reset(sk, skb);
645 discard:
646         if (opt_skb != NULL)
647                 __kfree_skb(opt_skb);
648         kfree_skb(skb);
649         return 0;
650
651 /* Handling IPV6_PKTOPTIONS skb the similar
652  * way it's done for net/ipv6/tcp_ipv6.c
653  */
654 ipv6_pktoptions:
655         if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
656                 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
657                         np->mcast_oif = inet6_iif(opt_skb);
658                 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
659                         np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
660                 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
661                         np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
662                 if (np->repflow)
663                         np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
664                 if (ipv6_opt_accepted(sk, opt_skb,
665                                       &DCCP_SKB_CB(opt_skb)->header.h6)) {
666                         skb_set_owner_r(opt_skb, sk);
667                         memmove(IP6CB(opt_skb),
668                                 &DCCP_SKB_CB(opt_skb)->header.h6,
669                                 sizeof(struct inet6_skb_parm));
670                         opt_skb = xchg(&np->pktoptions, opt_skb);
671                 } else {
672                         __kfree_skb(opt_skb);
673                         opt_skb = xchg(&np->pktoptions, NULL);
674                 }
675         }
676
677         kfree_skb(opt_skb);
678         return 0;
679 }
680
681 static int dccp_v6_rcv(struct sk_buff *skb)
682 {
683         const struct dccp_hdr *dh;
684         bool refcounted;
685         struct sock *sk;
686         int min_cov;
687
688         /* Step 1: Check header basics */
689
690         if (dccp_invalid_packet(skb))
691                 goto discard_it;
692
693         /* Step 1: If header checksum is incorrect, drop packet and return. */
694         if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
695                                      &ipv6_hdr(skb)->daddr)) {
696                 DCCP_WARN("dropped packet with invalid checksum\n");
697                 goto discard_it;
698         }
699
700         dh = dccp_hdr(skb);
701
702         DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
703         DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
704
705         if (dccp_packet_without_ack(skb))
706                 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
707         else
708                 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
709
710 lookup:
711         sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
712                                 dh->dccph_sport, dh->dccph_dport,
713                                 inet6_iif(skb), 0, &refcounted);
714         if (!sk) {
715                 dccp_pr_debug("failed to look up flow ID in table and "
716                               "get corresponding socket\n");
717                 goto no_dccp_socket;
718         }
719
720         /*
721          * Step 2:
722          *      ... or S.state == TIMEWAIT,
723          *              Generate Reset(No Connection) unless P.type == Reset
724          *              Drop packet and return
725          */
726         if (sk->sk_state == DCCP_TIME_WAIT) {
727                 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
728                 inet_twsk_put(inet_twsk(sk));
729                 goto no_dccp_socket;
730         }
731
732         if (sk->sk_state == DCCP_NEW_SYN_RECV) {
733                 struct request_sock *req = inet_reqsk(sk);
734                 struct sock *nsk;
735
736                 sk = req->rsk_listener;
737                 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
738                         inet_csk_reqsk_queue_drop_and_put(sk, req);
739                         goto lookup;
740                 }
741                 sock_hold(sk);
742                 refcounted = true;
743                 nsk = dccp_check_req(sk, skb, req);
744                 if (!nsk) {
745                         reqsk_put(req);
746                         goto discard_and_relse;
747                 }
748                 if (nsk == sk) {
749                         reqsk_put(req);
750                 } else if (dccp_child_process(sk, nsk, skb)) {
751                         dccp_v6_ctl_send_reset(sk, skb);
752                         goto discard_and_relse;
753                 } else {
754                         sock_put(sk);
755                         return 0;
756                 }
757         }
758         /*
759          * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
760          *      o if MinCsCov = 0, only packets with CsCov = 0 are accepted
761          *      o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
762          */
763         min_cov = dccp_sk(sk)->dccps_pcrlen;
764         if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
765                 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
766                               dh->dccph_cscov, min_cov);
767                 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
768                 goto discard_and_relse;
769         }
770
771         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
772                 goto discard_and_relse;
773
774         return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
775                                 refcounted) ? -1 : 0;
776
777 no_dccp_socket:
778         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
779                 goto discard_it;
780         /*
781          * Step 2:
782          *      If no socket ...
783          *              Generate Reset(No Connection) unless P.type == Reset
784          *              Drop packet and return
785          */
786         if (dh->dccph_type != DCCP_PKT_RESET) {
787                 DCCP_SKB_CB(skb)->dccpd_reset_code =
788                                         DCCP_RESET_CODE_NO_CONNECTION;
789                 dccp_v6_ctl_send_reset(sk, skb);
790         }
791
792 discard_it:
793         kfree_skb(skb);
794         return 0;
795
796 discard_and_relse:
797         if (refcounted)
798                 sock_put(sk);
799         goto discard_it;
800 }
801
802 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
803                            int addr_len)
804 {
805         struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
806         struct inet_connection_sock *icsk = inet_csk(sk);
807         struct inet_sock *inet = inet_sk(sk);
808         struct ipv6_pinfo *np = inet6_sk(sk);
809         struct dccp_sock *dp = dccp_sk(sk);
810         struct in6_addr *saddr = NULL, *final_p, final;
811         struct ipv6_txoptions *opt;
812         struct flowi6 fl6;
813         struct dst_entry *dst;
814         int addr_type;
815         int err;
816
817         dp->dccps_role = DCCP_ROLE_CLIENT;
818
819         if (addr_len < SIN6_LEN_RFC2133)
820                 return -EINVAL;
821
822         if (usin->sin6_family != AF_INET6)
823                 return -EAFNOSUPPORT;
824
825         memset(&fl6, 0, sizeof(fl6));
826
827         if (np->sndflow) {
828                 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
829                 IP6_ECN_flow_init(fl6.flowlabel);
830                 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
831                         struct ip6_flowlabel *flowlabel;
832                         flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
833                         if (flowlabel == NULL)
834                                 return -EINVAL;
835                         fl6_sock_release(flowlabel);
836                 }
837         }
838         /*
839          * connect() to INADDR_ANY means loopback (BSD'ism).
840          */
841         if (ipv6_addr_any(&usin->sin6_addr))
842                 usin->sin6_addr.s6_addr[15] = 1;
843
844         addr_type = ipv6_addr_type(&usin->sin6_addr);
845
846         if (addr_type & IPV6_ADDR_MULTICAST)
847                 return -ENETUNREACH;
848
849         if (addr_type & IPV6_ADDR_LINKLOCAL) {
850                 if (addr_len >= sizeof(struct sockaddr_in6) &&
851                     usin->sin6_scope_id) {
852                         /* If interface is set while binding, indices
853                          * must coincide.
854                          */
855                         if (sk->sk_bound_dev_if &&
856                             sk->sk_bound_dev_if != usin->sin6_scope_id)
857                                 return -EINVAL;
858
859                         sk->sk_bound_dev_if = usin->sin6_scope_id;
860                 }
861
862                 /* Connect to link-local address requires an interface */
863                 if (!sk->sk_bound_dev_if)
864                         return -EINVAL;
865         }
866
867         sk->sk_v6_daddr = usin->sin6_addr;
868         np->flow_label = fl6.flowlabel;
869
870         /*
871          * DCCP over IPv4
872          */
873         if (addr_type == IPV6_ADDR_MAPPED) {
874                 u32 exthdrlen = icsk->icsk_ext_hdr_len;
875                 struct sockaddr_in sin;
876
877                 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
878
879                 if (__ipv6_only_sock(sk))
880                         return -ENETUNREACH;
881
882                 sin.sin_family = AF_INET;
883                 sin.sin_port = usin->sin6_port;
884                 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
885
886                 icsk->icsk_af_ops = &dccp_ipv6_mapped;
887                 sk->sk_backlog_rcv = dccp_v4_do_rcv;
888
889                 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
890                 if (err) {
891                         icsk->icsk_ext_hdr_len = exthdrlen;
892                         icsk->icsk_af_ops = &dccp_ipv6_af_ops;
893                         sk->sk_backlog_rcv = dccp_v6_do_rcv;
894                         goto failure;
895                 }
896                 np->saddr = sk->sk_v6_rcv_saddr;
897                 return err;
898         }
899
900         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
901                 saddr = &sk->sk_v6_rcv_saddr;
902
903         fl6.flowi6_proto = IPPROTO_DCCP;
904         fl6.daddr = sk->sk_v6_daddr;
905         fl6.saddr = saddr ? *saddr : np->saddr;
906         fl6.flowi6_oif = sk->sk_bound_dev_if;
907         fl6.fl6_dport = usin->sin6_port;
908         fl6.fl6_sport = inet->inet_sport;
909         security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
910
911         opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
912         final_p = fl6_update_dst(&fl6, opt, &final);
913
914         dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
915         if (IS_ERR(dst)) {
916                 err = PTR_ERR(dst);
917                 goto failure;
918         }
919
920         if (saddr == NULL) {
921                 saddr = &fl6.saddr;
922                 sk->sk_v6_rcv_saddr = *saddr;
923         }
924
925         /* set the source address */
926         np->saddr = *saddr;
927         inet->inet_rcv_saddr = LOOPBACK4_IPV6;
928
929         ip6_dst_store(sk, dst, NULL, NULL);
930
931         icsk->icsk_ext_hdr_len = 0;
932         if (opt)
933                 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
934
935         inet->inet_dport = usin->sin6_port;
936
937         dccp_set_state(sk, DCCP_REQUESTING);
938         err = inet6_hash_connect(&dccp_death_row, sk);
939         if (err)
940                 goto late_failure;
941
942         dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
943                                                       sk->sk_v6_daddr.s6_addr32,
944                                                       inet->inet_sport,
945                                                       inet->inet_dport);
946         err = dccp_connect(sk);
947         if (err)
948                 goto late_failure;
949
950         return 0;
951
952 late_failure:
953         dccp_set_state(sk, DCCP_CLOSED);
954         __sk_dst_reset(sk);
955 failure:
956         inet->inet_dport = 0;
957         sk->sk_route_caps = 0;
958         return err;
959 }
960
961 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
962         .queue_xmit        = inet6_csk_xmit,
963         .send_check        = dccp_v6_send_check,
964         .rebuild_header    = inet6_sk_rebuild_header,
965         .conn_request      = dccp_v6_conn_request,
966         .syn_recv_sock     = dccp_v6_request_recv_sock,
967         .net_header_len    = sizeof(struct ipv6hdr),
968         .setsockopt        = ipv6_setsockopt,
969         .getsockopt        = ipv6_getsockopt,
970         .addr2sockaddr     = inet6_csk_addr2sockaddr,
971         .sockaddr_len      = sizeof(struct sockaddr_in6),
972 #ifdef CONFIG_COMPAT
973         .compat_setsockopt = compat_ipv6_setsockopt,
974         .compat_getsockopt = compat_ipv6_getsockopt,
975 #endif
976 };
977
978 /*
979  *      DCCP over IPv4 via INET6 API
980  */
981 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
982         .queue_xmit        = ip_queue_xmit,
983         .send_check        = dccp_v4_send_check,
984         .rebuild_header    = inet_sk_rebuild_header,
985         .conn_request      = dccp_v6_conn_request,
986         .syn_recv_sock     = dccp_v6_request_recv_sock,
987         .net_header_len    = sizeof(struct iphdr),
988         .setsockopt        = ipv6_setsockopt,
989         .getsockopt        = ipv6_getsockopt,
990         .addr2sockaddr     = inet6_csk_addr2sockaddr,
991         .sockaddr_len      = sizeof(struct sockaddr_in6),
992 #ifdef CONFIG_COMPAT
993         .compat_setsockopt = compat_ipv6_setsockopt,
994         .compat_getsockopt = compat_ipv6_getsockopt,
995 #endif
996 };
997
998 /* NOTE: A lot of things set to zero explicitly by call to
999  *       sk_alloc() so need not be done here.
1000  */
1001 static int dccp_v6_init_sock(struct sock *sk)
1002 {
1003         static __u8 dccp_v6_ctl_sock_initialized;
1004         int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1005
1006         if (err == 0) {
1007                 if (unlikely(!dccp_v6_ctl_sock_initialized))
1008                         dccp_v6_ctl_sock_initialized = 1;
1009                 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1010         }
1011
1012         return err;
1013 }
1014
1015 static void dccp_v6_destroy_sock(struct sock *sk)
1016 {
1017         dccp_destroy_sock(sk);
1018         inet6_destroy_sock(sk);
1019 }
1020
1021 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1022         .twsk_obj_size  = sizeof(struct dccp6_timewait_sock),
1023 };
1024
1025 static struct proto dccp_v6_prot = {
1026         .name              = "DCCPv6",
1027         .owner             = THIS_MODULE,
1028         .close             = dccp_close,
1029         .connect           = dccp_v6_connect,
1030         .disconnect        = dccp_disconnect,
1031         .ioctl             = dccp_ioctl,
1032         .init              = dccp_v6_init_sock,
1033         .setsockopt        = dccp_setsockopt,
1034         .getsockopt        = dccp_getsockopt,
1035         .sendmsg           = dccp_sendmsg,
1036         .recvmsg           = dccp_recvmsg,
1037         .backlog_rcv       = dccp_v6_do_rcv,
1038         .hash              = inet6_hash,
1039         .unhash            = inet_unhash,
1040         .accept            = inet_csk_accept,
1041         .get_port          = inet_csk_get_port,
1042         .shutdown          = dccp_shutdown,
1043         .destroy           = dccp_v6_destroy_sock,
1044         .orphan_count      = &dccp_orphan_count,
1045         .max_header        = MAX_DCCP_HEADER,
1046         .obj_size          = sizeof(struct dccp6_sock),
1047         .slab_flags        = SLAB_TYPESAFE_BY_RCU,
1048         .rsk_prot          = &dccp6_request_sock_ops,
1049         .twsk_prot         = &dccp6_timewait_sock_ops,
1050         .h.hashinfo        = &dccp_hashinfo,
1051 #ifdef CONFIG_COMPAT
1052         .compat_setsockopt = compat_dccp_setsockopt,
1053         .compat_getsockopt = compat_dccp_getsockopt,
1054 #endif
1055 };
1056
1057 static const struct inet6_protocol dccp_v6_protocol = {
1058         .handler        = dccp_v6_rcv,
1059         .err_handler    = dccp_v6_err,
1060         .flags          = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1061 };
1062
1063 static const struct proto_ops inet6_dccp_ops = {
1064         .family            = PF_INET6,
1065         .owner             = THIS_MODULE,
1066         .release           = inet6_release,
1067         .bind              = inet6_bind,
1068         .connect           = inet_stream_connect,
1069         .socketpair        = sock_no_socketpair,
1070         .accept            = inet_accept,
1071         .getname           = inet6_getname,
1072         .poll              = dccp_poll,
1073         .ioctl             = inet6_ioctl,
1074         .listen            = inet_dccp_listen,
1075         .shutdown          = inet_shutdown,
1076         .setsockopt        = sock_common_setsockopt,
1077         .getsockopt        = sock_common_getsockopt,
1078         .sendmsg           = inet_sendmsg,
1079         .recvmsg           = sock_common_recvmsg,
1080         .mmap              = sock_no_mmap,
1081         .sendpage          = sock_no_sendpage,
1082 #ifdef CONFIG_COMPAT
1083         .compat_setsockopt = compat_sock_common_setsockopt,
1084         .compat_getsockopt = compat_sock_common_getsockopt,
1085 #endif
1086 };
1087
1088 static struct inet_protosw dccp_v6_protosw = {
1089         .type           = SOCK_DCCP,
1090         .protocol       = IPPROTO_DCCP,
1091         .prot           = &dccp_v6_prot,
1092         .ops            = &inet6_dccp_ops,
1093         .flags          = INET_PROTOSW_ICSK,
1094 };
1095
1096 static int __net_init dccp_v6_init_net(struct net *net)
1097 {
1098         if (dccp_hashinfo.bhash == NULL)
1099                 return -ESOCKTNOSUPPORT;
1100
1101         return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1102                                     SOCK_DCCP, IPPROTO_DCCP, net);
1103 }
1104
1105 static void __net_exit dccp_v6_exit_net(struct net *net)
1106 {
1107         inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1108 }
1109
1110 static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list)
1111 {
1112         inet_twsk_purge(&dccp_hashinfo, AF_INET6);
1113 }
1114
1115 static struct pernet_operations dccp_v6_ops = {
1116         .init   = dccp_v6_init_net,
1117         .exit   = dccp_v6_exit_net,
1118         .exit_batch = dccp_v6_exit_batch,
1119 };
1120
1121 static int __init dccp_v6_init(void)
1122 {
1123         int err = proto_register(&dccp_v6_prot, 1);
1124
1125         if (err)
1126                 goto out;
1127
1128         inet6_register_protosw(&dccp_v6_protosw);
1129
1130         err = register_pernet_subsys(&dccp_v6_ops);
1131         if (err)
1132                 goto out_destroy_ctl_sock;
1133
1134         err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1135         if (err)
1136                 goto out_unregister_proto;
1137
1138 out:
1139         return err;
1140 out_unregister_proto:
1141         unregister_pernet_subsys(&dccp_v6_ops);
1142 out_destroy_ctl_sock:
1143         inet6_unregister_protosw(&dccp_v6_protosw);
1144         proto_unregister(&dccp_v6_prot);
1145         goto out;
1146 }
1147
1148 static void __exit dccp_v6_exit(void)
1149 {
1150         inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1151         unregister_pernet_subsys(&dccp_v6_ops);
1152         inet6_unregister_protosw(&dccp_v6_protosw);
1153         proto_unregister(&dccp_v6_prot);
1154 }
1155
1156 module_init(dccp_v6_init);
1157 module_exit(dccp_v6_exit);
1158
1159 /*
1160  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1161  * values directly, Also cover the case where the protocol is not specified,
1162  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1163  */
1164 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1165 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1166 MODULE_LICENSE("GPL");
1167 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1168 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");