?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
scsi: ufs: fix race between clock gating and devfreq scaling work
[linux/fpc-iii.git] / net / dccp / ipv6.c
blob6cbcf399d22b89d4286ab5a33a0448affbd2812b
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 /* add pseudo-header to DCCP checksum stored in skb->csum */
44 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
45 const struct in6_addr *saddr,
46 const struct in6_addr *daddr)
47 {
48 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
49 }
50
51 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
52 {
53 struct ipv6_pinfo *np = inet6_sk(sk);
54 struct dccp_hdr *dh = dccp_hdr(skb);
55
56 dccp_csum_outgoing(skb);
57 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
58 }
59
60 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
61 {
62 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
63 ipv6_hdr(skb)->saddr.s6_addr32,
64 dccp_hdr(skb)->dccph_dport,
65 dccp_hdr(skb)->dccph_sport );
66
67 }
68
69 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
70 u8 type, u8 code, int offset, __be32 info)
71 {
72 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
73 const struct dccp_hdr *dh;
74 struct dccp_sock *dp;
75 struct ipv6_pinfo *np;
76 struct sock *sk;
77 int err;
78 __u64 seq;
79 struct net *net = dev_net(skb->dev);
80
81 /* Only need dccph_dport & dccph_sport which are the first
82 * 4 bytes in dccp header.
83 * Our caller (icmpv6_notify()) already pulled 8 bytes for us.
84 */
85 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8);
86 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8);
87 dh = (struct dccp_hdr *)(skb->data + offset);
88
89 sk = __inet6_lookup_established(net, &dccp_hashinfo,
90 &hdr->daddr, dh->dccph_dport,
91 &hdr->saddr, ntohs(dh->dccph_sport),
92 inet6_iif(skb));
93
94 if (!sk) {
95 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
96 ICMP6_MIB_INERRORS);
97 return;
98 }
99
100 if (sk->sk_state == DCCP_TIME_WAIT) {
101 inet_twsk_put(inet_twsk(sk));
102 return;
103 }
104 seq = dccp_hdr_seq(dh);
105 if (sk->sk_state == DCCP_NEW_SYN_RECV)
106 return dccp_req_err(sk, seq);
107
108 bh_lock_sock(sk);
109 if (sock_owned_by_user(sk))
110 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
111
112 if (sk->sk_state == DCCP_CLOSED)
113 goto out;
114
115 dp = dccp_sk(sk);
116 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
117 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
118 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
119 goto out;
120 }
121
122 np = inet6_sk(sk);
123
124 if (type == NDISC_REDIRECT) {
125 if (!sock_owned_by_user(sk)) {
126 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
127
128 if (dst)
129 dst->ops->redirect(dst, sk, skb);
130 }
131 goto out;
132 }
133
134 if (type == ICMPV6_PKT_TOOBIG) {
135 struct dst_entry *dst = NULL;
136
137 if (!ip6_sk_accept_pmtu(sk))
138 goto out;
139
140 if (sock_owned_by_user(sk))
141 goto out;
142 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
143 goto out;
144
145 dst = inet6_csk_update_pmtu(sk, ntohl(info));
146 if (!dst)
147 goto out;
148
149 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
150 dccp_sync_mss(sk, dst_mtu(dst));
151 goto out;
152 }
153
154 icmpv6_err_convert(type, code, &err);
155
156 /* Might be for an request_sock */
157 switch (sk->sk_state) {
158 case DCCP_REQUESTING:
159 case DCCP_RESPOND: /* Cannot happen.
160 It can, it SYNs are crossed. --ANK */
161 if (!sock_owned_by_user(sk)) {
162 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
163 sk->sk_err = err;
164 /*
165 * Wake people up to see the error
166 * (see connect in sock.c)
167 */
168 sk->sk_error_report(sk);
169 dccp_done(sk);
170 } else
171 sk->sk_err_soft = err;
172 goto out;
173 }
174
175 if (!sock_owned_by_user(sk) && np->recverr) {
176 sk->sk_err = err;
177 sk->sk_error_report(sk);
178 } else
179 sk->sk_err_soft = err;
180
181 out:
182 bh_unlock_sock(sk);
183 sock_put(sk);
184 }
185
186
187 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
188 {
189 struct inet_request_sock *ireq = inet_rsk(req);
190 struct ipv6_pinfo *np = inet6_sk(sk);
191 struct sk_buff *skb;
192 struct in6_addr *final_p, final;
193 struct flowi6 fl6;
194 int err = -1;
195 struct dst_entry *dst;
196
197 memset(&fl6, 0, sizeof(fl6));
198 fl6.flowi6_proto = IPPROTO_DCCP;
199 fl6.daddr = ireq->ir_v6_rmt_addr;
200 fl6.saddr = ireq->ir_v6_loc_addr;
201 fl6.flowlabel = 0;
202 fl6.flowi6_oif = ireq->ir_iif;
203 fl6.fl6_dport = ireq->ir_rmt_port;
204 fl6.fl6_sport = htons(ireq->ir_num);
205 security_req_classify_flow(req, flowi6_to_flowi(&fl6));
206
207
208 rcu_read_lock();
209 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
210 rcu_read_unlock();
211
212 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
213 if (IS_ERR(dst)) {
214 err = PTR_ERR(dst);
215 dst = NULL;
216 goto done;
217 }
218
219 skb = dccp_make_response(sk, dst, req);
220 if (skb != NULL) {
221 struct dccp_hdr *dh = dccp_hdr(skb);
222 struct ipv6_txoptions *opt;
223
224 dh->dccph_checksum = dccp_v6_csum_finish(skb,
225 &ireq->ir_v6_loc_addr,
226 &ireq->ir_v6_rmt_addr);
227 fl6.daddr = ireq->ir_v6_rmt_addr;
228 rcu_read_lock();
229 opt = ireq->ipv6_opt;
230 if (!opt)
231 opt = rcu_dereference(np->opt);
232 err = ip6_xmit(sk, skb, &fl6, sk->sk_mark, opt, np->tclass);
233 rcu_read_unlock();
234 err = net_xmit_eval(err);
235 }
236
237 done:
238 dst_release(dst);
239 return err;
240 }
241
242 static void dccp_v6_reqsk_destructor(struct request_sock *req)
243 {
244 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
245 kfree(inet_rsk(req)->ipv6_opt);
246 kfree_skb(inet_rsk(req)->pktopts);
247 }
248
249 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
250 {
251 const struct ipv6hdr *rxip6h;
252 struct sk_buff *skb;
253 struct flowi6 fl6;
254 struct net *net = dev_net(skb_dst(rxskb)->dev);
255 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
256 struct dst_entry *dst;
257
258 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
259 return;
260
261 if (!ipv6_unicast_destination(rxskb))
262 return;
263
264 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
265 if (skb == NULL)
266 return;
267
268 rxip6h = ipv6_hdr(rxskb);
269 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
270 &rxip6h->daddr);
271
272 memset(&fl6, 0, sizeof(fl6));
273 fl6.daddr = rxip6h->saddr;
274 fl6.saddr = rxip6h->daddr;
275
276 fl6.flowi6_proto = IPPROTO_DCCP;
277 fl6.flowi6_oif = inet6_iif(rxskb);
278 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
279 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
280 security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
281
282 /* sk = NULL, but it is safe for now. RST socket required. */
283 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
284 if (!IS_ERR(dst)) {
285 skb_dst_set(skb, dst);
286 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0);
287 DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
288 DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
289 return;
290 }
291
292 kfree_skb(skb);
293 }
294
295 static struct request_sock_ops dccp6_request_sock_ops = {
296 .family = AF_INET6,
297 .obj_size = sizeof(struct dccp6_request_sock),
298 .rtx_syn_ack = dccp_v6_send_response,
299 .send_ack = dccp_reqsk_send_ack,
300 .destructor = dccp_v6_reqsk_destructor,
301 .send_reset = dccp_v6_ctl_send_reset,
302 .syn_ack_timeout = dccp_syn_ack_timeout,
303 };
304
305 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
306 {
307 struct request_sock *req;
308 struct dccp_request_sock *dreq;
309 struct inet_request_sock *ireq;
310 struct ipv6_pinfo *np = inet6_sk(sk);
311 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
312 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
313
314 if (skb->protocol == htons(ETH_P_IP))
315 return dccp_v4_conn_request(sk, skb);
316
317 if (!ipv6_unicast_destination(skb))
318 return 0; /* discard, don't send a reset here */
319
320 if (dccp_bad_service_code(sk, service)) {
321 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
322 goto drop;
323 }
324 /*
325 * There are no SYN attacks on IPv6, yet...
326 */
327 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
328 if (inet_csk_reqsk_queue_is_full(sk))
329 goto drop;
330
331 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
332 goto drop;
333
334 req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
335 if (req == NULL)
336 goto drop;
337
338 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
339 goto drop_and_free;
340
341 dreq = dccp_rsk(req);
342 if (dccp_parse_options(sk, dreq, skb))
343 goto drop_and_free;
344
345 if (security_inet_conn_request(sk, skb, req))
346 goto drop_and_free;
347
348 ireq = inet_rsk(req);
349 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
350 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
351 ireq->ireq_family = AF_INET6;
352 ireq->ir_mark = inet_request_mark(sk, skb);
353
354 if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
355 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
356 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
357 atomic_inc(&skb->users);
358 ireq->pktopts = skb;
359 }
360 ireq->ir_iif = sk->sk_bound_dev_if;
361
362 /* So that link locals have meaning */
363 if (!sk->sk_bound_dev_if &&
364 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
365 ireq->ir_iif = inet6_iif(skb);
366
367 /*
368 * Step 3: Process LISTEN state
369 *
370 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
371 *
372 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
373 */
374 dreq->dreq_isr = dcb->dccpd_seq;
375 dreq->dreq_gsr = dreq->dreq_isr;
376 dreq->dreq_iss = dccp_v6_init_sequence(skb);
377 dreq->dreq_gss = dreq->dreq_iss;
378 dreq->dreq_service = service;
379
380 if (dccp_v6_send_response(sk, req))
381 goto drop_and_free;
382
383 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
384 reqsk_put(req);
385 return 0;
386
387 drop_and_free:
388 reqsk_free(req);
389 drop:
390 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
391 return -1;
392 }
393
394 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
395 struct sk_buff *skb,
396 struct request_sock *req,
397 struct dst_entry *dst,
398 struct request_sock *req_unhash,
399 bool *own_req)
400 {
401 struct inet_request_sock *ireq = inet_rsk(req);
402 struct ipv6_pinfo *newnp;
403 const struct ipv6_pinfo *np = inet6_sk(sk);
404 struct ipv6_txoptions *opt;
405 struct inet_sock *newinet;
406 struct dccp6_sock *newdp6;
407 struct sock *newsk;
408
409 if (skb->protocol == htons(ETH_P_IP)) {
410 /*
411 * v6 mapped
412 */
413 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
414 req_unhash, own_req);
415 if (newsk == NULL)
416 return NULL;
417
418 newdp6 = (struct dccp6_sock *)newsk;
419 newinet = inet_sk(newsk);
420 newinet->pinet6 = &newdp6->inet6;
421 newnp = inet6_sk(newsk);
422
423 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
424
425 newnp->saddr = newsk->sk_v6_rcv_saddr;
426
427 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
428 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
429 newnp->pktoptions = NULL;
430 newnp->opt = NULL;
431 newnp->ipv6_mc_list = NULL;
432 newnp->ipv6_ac_list = NULL;
433 newnp->ipv6_fl_list = NULL;
434 newnp->mcast_oif = inet6_iif(skb);
435 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
436
437 /*
438 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
439 * here, dccp_create_openreq_child now does this for us, see the comment in
440 * that function for the gory details. -acme
441 */
442
443 /* It is tricky place. Until this moment IPv4 tcp
444 worked with IPv6 icsk.icsk_af_ops.
445 Sync it now.
446 */
447 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
448
449 return newsk;
450 }
451
452
453 if (sk_acceptq_is_full(sk))
454 goto out_overflow;
455
456 if (!dst) {
457 struct flowi6 fl6;
458
459 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
460 if (!dst)
461 goto out;
462 }
463
464 newsk = dccp_create_openreq_child(sk, req, skb);
465 if (newsk == NULL)
466 goto out_nonewsk;
467
468 /*
469 * No need to charge this sock to the relevant IPv6 refcnt debug socks
470 * count here, dccp_create_openreq_child now does this for us, see the
471 * comment in that function for the gory details. -acme
472 */
473
474 ip6_dst_store(newsk, dst, NULL, NULL);
475 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
476 NETIF_F_TSO);
477 newdp6 = (struct dccp6_sock *)newsk;
478 newinet = inet_sk(newsk);
479 newinet->pinet6 = &newdp6->inet6;
480 newnp = inet6_sk(newsk);
481
482 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
483
484 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
485 newnp->saddr = ireq->ir_v6_loc_addr;
486 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
487 newsk->sk_bound_dev_if = ireq->ir_iif;
488
489 /* Now IPv6 options...
490
491 First: no IPv4 options.
492 */
493 newinet->inet_opt = NULL;
494
495 /* Clone RX bits */
496 newnp->rxopt.all = np->rxopt.all;
497
498 newnp->ipv6_mc_list = NULL;
499 newnp->ipv6_ac_list = NULL;
500 newnp->ipv6_fl_list = NULL;
501 newnp->pktoptions = NULL;
502 newnp->opt = NULL;
503 newnp->mcast_oif = inet6_iif(skb);
504 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
505
506 /*
507 * Clone native IPv6 options from listening socket (if any)
508 *
509 * Yes, keeping reference count would be much more clever, but we make
510 * one more one thing there: reattach optmem to newsk.
511 */
512 opt = ireq->ipv6_opt;
513 if (!opt)
514 opt = rcu_dereference(np->opt);
515 if (opt) {
516 opt = ipv6_dup_options(newsk, opt);
517 RCU_INIT_POINTER(newnp->opt, opt);
518 }
519 inet_csk(newsk)->icsk_ext_hdr_len = 0;
520 if (opt)
521 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
522 opt->opt_flen;
523
524 dccp_sync_mss(newsk, dst_mtu(dst));
525
526 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
527 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
528
529 if (__inet_inherit_port(sk, newsk) < 0) {
530 inet_csk_prepare_forced_close(newsk);
531 dccp_done(newsk);
532 goto out;
533 }
534 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
535 /* Clone pktoptions received with SYN, if we own the req */
536 if (*own_req && ireq->pktopts) {
537 newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
538 consume_skb(ireq->pktopts);
539 ireq->pktopts = NULL;
540 if (newnp->pktoptions)
541 skb_set_owner_r(newnp->pktoptions, newsk);
542 }
543
544 return newsk;
545
546 out_overflow:
547 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
548 out_nonewsk:
549 dst_release(dst);
550 out:
551 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
552 return NULL;
553 }
554
555 /* The socket must have it's spinlock held when we get
556 * here.
557 *
558 * We have a potential double-lock case here, so even when
559 * doing backlog processing we use the BH locking scheme.
560 * This is because we cannot sleep with the original spinlock
561 * held.
562 */
563 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
564 {
565 struct ipv6_pinfo *np = inet6_sk(sk);
566 struct sk_buff *opt_skb = NULL;
567
568 /* Imagine: socket is IPv6. IPv4 packet arrives,
569 goes to IPv4 receive handler and backlogged.
570 From backlog it always goes here. Kerboom...
571 Fortunately, dccp_rcv_established and rcv_established
572 handle them correctly, but it is not case with
573 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
574 */
575
576 if (skb->protocol == htons(ETH_P_IP))
577 return dccp_v4_do_rcv(sk, skb);
578
579 if (sk_filter(sk, skb))
580 goto discard;
581
582 /*
583 * socket locking is here for SMP purposes as backlog rcv is currently
584 * called with bh processing disabled.
585 */
586
587 /* Do Stevens' IPV6_PKTOPTIONS.
588
589 Yes, guys, it is the only place in our code, where we
590 may make it not affecting IPv4.
591 The rest of code is protocol independent,
592 and I do not like idea to uglify IPv4.
593
594 Actually, all the idea behind IPV6_PKTOPTIONS
595 looks not very well thought. For now we latch
596 options, received in the last packet, enqueued
597 by tcp. Feel free to propose better solution.
598 --ANK (980728)
599 */
600 if (np->rxopt.all)
601 /*
602 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
603 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
604 */
605 opt_skb = skb_clone(skb, GFP_ATOMIC);
606
607 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
608 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
609 goto reset;
610 if (opt_skb) {
611 /* XXX This is where we would goto ipv6_pktoptions. */
612 __kfree_skb(opt_skb);
613 }
614 return 0;
615 }
616
617 /*
618 * Step 3: Process LISTEN state
619 * If S.state == LISTEN,
620 * If P.type == Request or P contains a valid Init Cookie option,
621 * (* Must scan the packet's options to check for Init
622 * Cookies. Only Init Cookies are processed here,
623 * however; other options are processed in Step 8. This
624 * scan need only be performed if the endpoint uses Init
625 * Cookies *)
626 * (* Generate a new socket and switch to that socket *)
627 * Set S := new socket for this port pair
628 * S.state = RESPOND
629 * Choose S.ISS (initial seqno) or set from Init Cookies
630 * Initialize S.GAR := S.ISS
631 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
632 * Continue with S.state == RESPOND
633 * (* A Response packet will be generated in Step 11 *)
634 * Otherwise,
635 * Generate Reset(No Connection) unless P.type == Reset
636 * Drop packet and return
637 *
638 * NOTE: the check for the packet types is done in
639 * dccp_rcv_state_process
640 */
641
642 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
643 goto reset;
644 if (opt_skb) {
645 /* XXX This is where we would goto ipv6_pktoptions. */
646 __kfree_skb(opt_skb);
647 }
648 return 0;
649
650 reset:
651 dccp_v6_ctl_send_reset(sk, skb);
652 discard:
653 if (opt_skb != NULL)
654 __kfree_skb(opt_skb);
655 kfree_skb(skb);
656 return 0;
657 }
658
659 static int dccp_v6_rcv(struct sk_buff *skb)
660 {
661 const struct dccp_hdr *dh;
662 bool refcounted;
663 struct sock *sk;
664 int min_cov;
665
666 /* Step 1: Check header basics */
667
668 if (dccp_invalid_packet(skb))
669 goto discard_it;
670
671 /* Step 1: If header checksum is incorrect, drop packet and return. */
672 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
673 &ipv6_hdr(skb)->daddr)) {
674 DCCP_WARN("dropped packet with invalid checksum\n");
675 goto discard_it;
676 }
677
678 dh = dccp_hdr(skb);
679
680 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
681 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
682
683 if (dccp_packet_without_ack(skb))
684 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
685 else
686 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
687
688 lookup:
689 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
690 dh->dccph_sport, dh->dccph_dport,
691 inet6_iif(skb), &refcounted);
692 if (!sk) {
693 dccp_pr_debug("failed to look up flow ID in table and "
694 "get corresponding socket\n");
695 goto no_dccp_socket;
696 }
697
698 /*
699 * Step 2:
700 * ... or S.state == TIMEWAIT,
701 * Generate Reset(No Connection) unless P.type == Reset
702 * Drop packet and return
703 */
704 if (sk->sk_state == DCCP_TIME_WAIT) {
705 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
706 inet_twsk_put(inet_twsk(sk));
707 goto no_dccp_socket;
708 }
709
710 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
711 struct request_sock *req = inet_reqsk(sk);
712 struct sock *nsk;
713
714 sk = req->rsk_listener;
715 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
716 inet_csk_reqsk_queue_drop_and_put(sk, req);
717 goto lookup;
718 }
719 sock_hold(sk);
720 refcounted = true;
721 nsk = dccp_check_req(sk, skb, req);
722 if (!nsk) {
723 reqsk_put(req);
724 goto discard_and_relse;
725 }
726 if (nsk == sk) {
727 reqsk_put(req);
728 } else if (dccp_child_process(sk, nsk, skb)) {
729 dccp_v6_ctl_send_reset(sk, skb);
730 goto discard_and_relse;
731 } else {
732 sock_put(sk);
733 return 0;
734 }
735 }
736 /*
737 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
738 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
739 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
740 */
741 min_cov = dccp_sk(sk)->dccps_pcrlen;
742 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
743 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
744 dh->dccph_cscov, min_cov);
745 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
746 goto discard_and_relse;
747 }
748
749 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
750 goto discard_and_relse;
751
752 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
753 refcounted) ? -1 : 0;
754
755 no_dccp_socket:
756 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
757 goto discard_it;
758 /*
759 * Step 2:
760 * If no socket ...
761 * Generate Reset(No Connection) unless P.type == Reset
762 * Drop packet and return
763 */
764 if (dh->dccph_type != DCCP_PKT_RESET) {
765 DCCP_SKB_CB(skb)->dccpd_reset_code =
766 DCCP_RESET_CODE_NO_CONNECTION;
767 dccp_v6_ctl_send_reset(sk, skb);
768 }
769
770 discard_it:
771 kfree_skb(skb);
772 return 0;
773
774 discard_and_relse:
775 if (refcounted)
776 sock_put(sk);
777 goto discard_it;
778 }
779
780 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
781 int addr_len)
782 {
783 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
784 struct inet_connection_sock *icsk = inet_csk(sk);
785 struct inet_sock *inet = inet_sk(sk);
786 struct ipv6_pinfo *np = inet6_sk(sk);
787 struct dccp_sock *dp = dccp_sk(sk);
788 struct in6_addr *saddr = NULL, *final_p, final;
789 struct ipv6_txoptions *opt;
790 struct flowi6 fl6;
791 struct dst_entry *dst;
792 int addr_type;
793 int err;
794
795 dp->dccps_role = DCCP_ROLE_CLIENT;
796
797 if (addr_len < SIN6_LEN_RFC2133)
798 return -EINVAL;
799
800 if (usin->sin6_family != AF_INET6)
801 return -EAFNOSUPPORT;
802
803 memset(&fl6, 0, sizeof(fl6));
804
805 if (np->sndflow) {
806 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
807 IP6_ECN_flow_init(fl6.flowlabel);
808 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
809 struct ip6_flowlabel *flowlabel;
810 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
811 if (flowlabel == NULL)
812 return -EINVAL;
813 fl6_sock_release(flowlabel);
814 }
815 }
816 /*
817 * connect() to INADDR_ANY means loopback (BSD'ism).
818 */
819 if (ipv6_addr_any(&usin->sin6_addr))
820 usin->sin6_addr.s6_addr[15] = 1;
821
822 addr_type = ipv6_addr_type(&usin->sin6_addr);
823
824 if (addr_type & IPV6_ADDR_MULTICAST)
825 return -ENETUNREACH;
826
827 if (addr_type & IPV6_ADDR_LINKLOCAL) {
828 if (addr_len >= sizeof(struct sockaddr_in6) &&
829 usin->sin6_scope_id) {
830 /* If interface is set while binding, indices
831 * must coincide.
832 */
833 if (sk->sk_bound_dev_if &&
834 sk->sk_bound_dev_if != usin->sin6_scope_id)
835 return -EINVAL;
836
837 sk->sk_bound_dev_if = usin->sin6_scope_id;
838 }
839
840 /* Connect to link-local address requires an interface */
841 if (!sk->sk_bound_dev_if)
842 return -EINVAL;
843 }
844
845 sk->sk_v6_daddr = usin->sin6_addr;
846 np->flow_label = fl6.flowlabel;
847
848 /*
849 * DCCP over IPv4
850 */
851 if (addr_type == IPV6_ADDR_MAPPED) {
852 u32 exthdrlen = icsk->icsk_ext_hdr_len;
853 struct sockaddr_in sin;
854
855 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
856
857 if (__ipv6_only_sock(sk))
858 return -ENETUNREACH;
859
860 sin.sin_family = AF_INET;
861 sin.sin_port = usin->sin6_port;
862 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
863
864 icsk->icsk_af_ops = &dccp_ipv6_mapped;
865 sk->sk_backlog_rcv = dccp_v4_do_rcv;
866
867 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
868 if (err) {
869 icsk->icsk_ext_hdr_len = exthdrlen;
870 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
871 sk->sk_backlog_rcv = dccp_v6_do_rcv;
872 goto failure;
873 }
874 np->saddr = sk->sk_v6_rcv_saddr;
875 return err;
876 }
877
878 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
879 saddr = &sk->sk_v6_rcv_saddr;
880
881 fl6.flowi6_proto = IPPROTO_DCCP;
882 fl6.daddr = sk->sk_v6_daddr;
883 fl6.saddr = saddr ? *saddr : np->saddr;
884 fl6.flowi6_oif = sk->sk_bound_dev_if;
885 fl6.fl6_dport = usin->sin6_port;
886 fl6.fl6_sport = inet->inet_sport;
887 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
888
889 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
890 final_p = fl6_update_dst(&fl6, opt, &final);
891
892 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
893 if (IS_ERR(dst)) {
894 err = PTR_ERR(dst);
895 goto failure;
896 }
897
898 if (saddr == NULL) {
899 saddr = &fl6.saddr;
900 sk->sk_v6_rcv_saddr = *saddr;
901 }
902
903 /* set the source address */
904 np->saddr = *saddr;
905 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
906
907 ip6_dst_store(sk, dst, NULL, NULL);
908
909 icsk->icsk_ext_hdr_len = 0;
910 if (opt)
911 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
912
913 inet->inet_dport = usin->sin6_port;
914
915 dccp_set_state(sk, DCCP_REQUESTING);
916 err = inet6_hash_connect(&dccp_death_row, sk);
917 if (err)
918 goto late_failure;
919
920 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
921 sk->sk_v6_daddr.s6_addr32,
922 inet->inet_sport,
923 inet->inet_dport);
924 err = dccp_connect(sk);
925 if (err)
926 goto late_failure;
927
928 return 0;
929
930 late_failure:
931 dccp_set_state(sk, DCCP_CLOSED);
932 __sk_dst_reset(sk);
933 failure:
934 inet->inet_dport = 0;
935 sk->sk_route_caps = 0;
936 return err;
937 }
938
939 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
940 .queue_xmit = inet6_csk_xmit,
941 .send_check = dccp_v6_send_check,
942 .rebuild_header = inet6_sk_rebuild_header,
943 .conn_request = dccp_v6_conn_request,
944 .syn_recv_sock = dccp_v6_request_recv_sock,
945 .net_header_len = sizeof(struct ipv6hdr),
946 .setsockopt = ipv6_setsockopt,
947 .getsockopt = ipv6_getsockopt,
948 .addr2sockaddr = inet6_csk_addr2sockaddr,
949 .sockaddr_len = sizeof(struct sockaddr_in6),
950 .bind_conflict = inet6_csk_bind_conflict,
951 #ifdef CONFIG_COMPAT
952 .compat_setsockopt = compat_ipv6_setsockopt,
953 .compat_getsockopt = compat_ipv6_getsockopt,
954 #endif
955 };
956
957 /*
958 * DCCP over IPv4 via INET6 API
959 */
960 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
961 .queue_xmit = ip_queue_xmit,
962 .send_check = dccp_v4_send_check,
963 .rebuild_header = inet_sk_rebuild_header,
964 .conn_request = dccp_v6_conn_request,
965 .syn_recv_sock = dccp_v6_request_recv_sock,
966 .net_header_len = sizeof(struct iphdr),
967 .setsockopt = ipv6_setsockopt,
968 .getsockopt = ipv6_getsockopt,
969 .addr2sockaddr = inet6_csk_addr2sockaddr,
970 .sockaddr_len = sizeof(struct sockaddr_in6),
971 .bind_conflict = inet6_csk_bind_conflict,
972 #ifdef CONFIG_COMPAT
973 .compat_setsockopt = compat_ipv6_setsockopt,
974 .compat_getsockopt = compat_ipv6_getsockopt,
975 #endif
976 };
977
978 /* NOTE: A lot of things set to zero explicitly by call to
979 * sk_alloc() so need not be done here.
980 */
981 static int dccp_v6_init_sock(struct sock *sk)
982 {
983 static __u8 dccp_v6_ctl_sock_initialized;
984 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
985
986 if (err == 0) {
987 if (unlikely(!dccp_v6_ctl_sock_initialized))
988 dccp_v6_ctl_sock_initialized = 1;
989 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
990 }
991
992 return err;
993 }
994
995 static void dccp_v6_destroy_sock(struct sock *sk)
996 {
997 dccp_destroy_sock(sk);
998 inet6_destroy_sock(sk);
999 }
1000
1001 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1002 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1003 };
1004
1005 static struct proto dccp_v6_prot = {
1006 .name = "DCCPv6",
1007 .owner = THIS_MODULE,
1008 .close = dccp_close,
1009 .connect = dccp_v6_connect,
1010 .disconnect = dccp_disconnect,
1011 .ioctl = dccp_ioctl,
1012 .init = dccp_v6_init_sock,
1013 .setsockopt = dccp_setsockopt,
1014 .getsockopt = dccp_getsockopt,
1015 .sendmsg = dccp_sendmsg,
1016 .recvmsg = dccp_recvmsg,
1017 .backlog_rcv = dccp_v6_do_rcv,
1018 .hash = inet6_hash,
1019 .unhash = inet_unhash,
1020 .accept = inet_csk_accept,
1021 .get_port = inet_csk_get_port,
1022 .shutdown = dccp_shutdown,
1023 .destroy = dccp_v6_destroy_sock,
1024 .orphan_count = &dccp_orphan_count,
1025 .max_header = MAX_DCCP_HEADER,
1026 .obj_size = sizeof(struct dccp6_sock),
1027 .slab_flags = SLAB_DESTROY_BY_RCU,
1028 .rsk_prot = &dccp6_request_sock_ops,
1029 .twsk_prot = &dccp6_timewait_sock_ops,
1030 .h.hashinfo = &dccp_hashinfo,
1031 #ifdef CONFIG_COMPAT
1032 .compat_setsockopt = compat_dccp_setsockopt,
1033 .compat_getsockopt = compat_dccp_getsockopt,
1034 #endif
1035 };
1036
1037 static const struct inet6_protocol dccp_v6_protocol = {
1038 .handler = dccp_v6_rcv,
1039 .err_handler = dccp_v6_err,
1040 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1041 };
1042
1043 static const struct proto_ops inet6_dccp_ops = {
1044 .family = PF_INET6,
1045 .owner = THIS_MODULE,
1046 .release = inet6_release,
1047 .bind = inet6_bind,
1048 .connect = inet_stream_connect,
1049 .socketpair = sock_no_socketpair,
1050 .accept = inet_accept,
1051 .getname = inet6_getname,
1052 .poll = dccp_poll,
1053 .ioctl = inet6_ioctl,
1054 .listen = inet_dccp_listen,
1055 .shutdown = inet_shutdown,
1056 .setsockopt = sock_common_setsockopt,
1057 .getsockopt = sock_common_getsockopt,
1058 .sendmsg = inet_sendmsg,
1059 .recvmsg = sock_common_recvmsg,
1060 .mmap = sock_no_mmap,
1061 .sendpage = sock_no_sendpage,
1062 #ifdef CONFIG_COMPAT
1063 .compat_setsockopt = compat_sock_common_setsockopt,
1064 .compat_getsockopt = compat_sock_common_getsockopt,
1065 #endif
1066 };
1067
1068 static struct inet_protosw dccp_v6_protosw = {
1069 .type = SOCK_DCCP,
1070 .protocol = IPPROTO_DCCP,
1071 .prot = &dccp_v6_prot,
1072 .ops = &inet6_dccp_ops,
1073 .flags = INET_PROTOSW_ICSK,
1074 };
1075
1076 static int __net_init dccp_v6_init_net(struct net *net)
1077 {
1078 if (dccp_hashinfo.bhash == NULL)
1079 return -ESOCKTNOSUPPORT;
1080
1081 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1082 SOCK_DCCP, IPPROTO_DCCP, net);
1083 }
1084
1085 static void __net_exit dccp_v6_exit_net(struct net *net)
1086 {
1087 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1088 }
1089
1090 static struct pernet_operations dccp_v6_ops = {
1091 .init = dccp_v6_init_net,
1092 .exit = dccp_v6_exit_net,
1093 };
1094
1095 static int __init dccp_v6_init(void)
1096 {
1097 int err = proto_register(&dccp_v6_prot, 1);
1098
1099 if (err != 0)
1100 goto out;
1101
1102 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1103 if (err != 0)
1104 goto out_unregister_proto;
1105
1106 inet6_register_protosw(&dccp_v6_protosw);
1107
1108 err = register_pernet_subsys(&dccp_v6_ops);
1109 if (err != 0)
1110 goto out_destroy_ctl_sock;
1111 out:
1112 return err;
1113
1114 out_destroy_ctl_sock:
1115 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1116 inet6_unregister_protosw(&dccp_v6_protosw);
1117 out_unregister_proto:
1118 proto_unregister(&dccp_v6_prot);
1119 goto out;
1120 }
1121
1122 static void __exit dccp_v6_exit(void)
1123 {
1124 unregister_pernet_subsys(&dccp_v6_ops);
1125 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1126 inet6_unregister_protosw(&dccp_v6_protosw);
1127 proto_unregister(&dccp_v6_prot);
1128 }
1129
1130 module_init(dccp_v6_init);
1131 module_exit(dccp_v6_exit);
1132
1133 /*
1134 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1135 * values directly, Also cover the case where the protocol is not specified,
1136 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1137 */
1138 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1139 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1140 MODULE_LICENSE("GPL");
1141 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1142 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
Linux with added FPC-III support