Linux 2.6.20.7
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / dccp / ipv6.c
blob6b91a9dd04111298c96127ea49e18d864602e9f9
1 /*
2 * DCCP over IPv6
3 * Linux INET6 implementation
5 * Based on net/dccp6/ipv6.c
7 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
15 #include <linux/module.h>
16 #include <linux/random.h>
17 #include <linux/xfrm.h>
19 #include <net/addrconf.h>
20 #include <net/inet_common.h>
21 #include <net/inet_hashtables.h>
22 #include <net/inet_sock.h>
23 #include <net/inet6_connection_sock.h>
24 #include <net/inet6_hashtables.h>
25 #include <net/ip6_route.h>
26 #include <net/ipv6.h>
27 #include <net/protocol.h>
28 #include <net/transp_v6.h>
29 #include <net/ip6_checksum.h>
30 #include <net/xfrm.h>
32 #include "dccp.h"
33 #include "ipv6.h"
34 #include "feat.h"
36 /* Socket used for sending RSTs and ACKs */
37 static struct socket *dccp_v6_ctl_socket;
39 static struct inet_connection_sock_af_ops dccp_ipv6_mapped;
40 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
42 static int dccp_v6_get_port(struct sock *sk, unsigned short snum)
44 return inet_csk_get_port(&dccp_hashinfo, sk, snum,
45 inet6_csk_bind_conflict);
48 static void dccp_v6_hash(struct sock *sk)
50 if (sk->sk_state != DCCP_CLOSED) {
51 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
52 dccp_hash(sk);
53 return;
55 local_bh_disable();
56 __inet6_hash(&dccp_hashinfo, sk);
57 local_bh_enable();
61 /* add pseudo-header to DCCP checksum stored in skb->csum */
62 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
63 struct in6_addr *saddr,
64 struct in6_addr *daddr)
66 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
69 static inline void dccp_v6_send_check(struct sock *sk, int unused_value,
70 struct sk_buff *skb)
72 struct ipv6_pinfo *np = inet6_sk(sk);
73 struct dccp_hdr *dh = dccp_hdr(skb);
75 dccp_csum_outgoing(skb);
76 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr);
79 static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
80 __be16 sport, __be16 dport )
82 return secure_tcpv6_sequence_number(saddr, daddr, sport, dport);
85 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb)
87 return secure_dccpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
88 skb->nh.ipv6h->saddr.s6_addr32,
89 dccp_hdr(skb)->dccph_dport,
90 dccp_hdr(skb)->dccph_sport );
94 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
95 int type, int code, int offset, __be32 info)
97 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
98 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
99 struct ipv6_pinfo *np;
100 struct sock *sk;
101 int err;
102 __u64 seq;
104 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
105 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
107 if (sk == NULL) {
108 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
109 return;
112 if (sk->sk_state == DCCP_TIME_WAIT) {
113 inet_twsk_put(inet_twsk(sk));
114 return;
117 bh_lock_sock(sk);
118 if (sock_owned_by_user(sk))
119 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
121 if (sk->sk_state == DCCP_CLOSED)
122 goto out;
124 np = inet6_sk(sk);
126 if (type == ICMPV6_PKT_TOOBIG) {
127 struct dst_entry *dst = NULL;
129 if (sock_owned_by_user(sk))
130 goto out;
131 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
132 goto out;
134 /* icmp should have updated the destination cache entry */
135 dst = __sk_dst_check(sk, np->dst_cookie);
136 if (dst == NULL) {
137 struct inet_sock *inet = inet_sk(sk);
138 struct flowi fl;
140 /* BUGGG_FUTURE: Again, it is not clear how
141 to handle rthdr case. Ignore this complexity
142 for now.
144 memset(&fl, 0, sizeof(fl));
145 fl.proto = IPPROTO_DCCP;
146 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
147 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
148 fl.oif = sk->sk_bound_dev_if;
149 fl.fl_ip_dport = inet->dport;
150 fl.fl_ip_sport = inet->sport;
151 security_sk_classify_flow(sk, &fl);
153 err = ip6_dst_lookup(sk, &dst, &fl);
154 if (err) {
155 sk->sk_err_soft = -err;
156 goto out;
159 err = xfrm_lookup(&dst, &fl, sk, 0);
160 if (err < 0) {
161 sk->sk_err_soft = -err;
162 goto out;
164 } else
165 dst_hold(dst);
167 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
168 dccp_sync_mss(sk, dst_mtu(dst));
169 } /* else let the usual retransmit timer handle it */
170 dst_release(dst);
171 goto out;
174 icmpv6_err_convert(type, code, &err);
176 seq = DCCP_SKB_CB(skb)->dccpd_seq;
177 /* Might be for an request_sock */
178 switch (sk->sk_state) {
179 struct request_sock *req, **prev;
180 case DCCP_LISTEN:
181 if (sock_owned_by_user(sk))
182 goto out;
184 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
185 &hdr->daddr, &hdr->saddr,
186 inet6_iif(skb));
187 if (req == NULL)
188 goto out;
191 * ICMPs are not backlogged, hence we cannot get an established
192 * socket here.
194 BUG_TRAP(req->sk == NULL);
196 if (seq != dccp_rsk(req)->dreq_iss) {
197 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
198 goto out;
201 inet_csk_reqsk_queue_drop(sk, req, prev);
202 goto out;
204 case DCCP_REQUESTING:
205 case DCCP_RESPOND: /* Cannot happen.
206 It can, it SYNs are crossed. --ANK */
207 if (!sock_owned_by_user(sk)) {
208 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
209 sk->sk_err = err;
211 * Wake people up to see the error
212 * (see connect in sock.c)
214 sk->sk_error_report(sk);
215 dccp_done(sk);
216 } else
217 sk->sk_err_soft = err;
218 goto out;
221 if (!sock_owned_by_user(sk) && np->recverr) {
222 sk->sk_err = err;
223 sk->sk_error_report(sk);
224 } else
225 sk->sk_err_soft = err;
227 out:
228 bh_unlock_sock(sk);
229 sock_put(sk);
233 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
234 struct dst_entry *dst)
236 struct inet6_request_sock *ireq6 = inet6_rsk(req);
237 struct ipv6_pinfo *np = inet6_sk(sk);
238 struct sk_buff *skb;
239 struct ipv6_txoptions *opt = NULL;
240 struct in6_addr *final_p = NULL, final;
241 struct flowi fl;
242 int err = -1;
244 memset(&fl, 0, sizeof(fl));
245 fl.proto = IPPROTO_DCCP;
246 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
247 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
248 fl.fl6_flowlabel = 0;
249 fl.oif = ireq6->iif;
250 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
251 fl.fl_ip_sport = inet_sk(sk)->sport;
252 security_req_classify_flow(req, &fl);
254 if (dst == NULL) {
255 opt = np->opt;
256 if (opt == NULL &&
257 np->rxopt.bits.osrcrt == 2 &&
258 ireq6->pktopts) {
259 struct sk_buff *pktopts = ireq6->pktopts;
260 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
262 if (rxopt->srcrt)
263 opt = ipv6_invert_rthdr(sk,
264 (struct ipv6_rt_hdr *)(pktopts->nh.raw +
265 rxopt->srcrt));
268 if (opt != NULL && opt->srcrt != NULL) {
269 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
271 ipv6_addr_copy(&final, &fl.fl6_dst);
272 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
273 final_p = &final;
276 err = ip6_dst_lookup(sk, &dst, &fl);
277 if (err)
278 goto done;
280 if (final_p)
281 ipv6_addr_copy(&fl.fl6_dst, final_p);
283 err = xfrm_lookup(&dst, &fl, sk, 0);
284 if (err < 0)
285 goto done;
288 skb = dccp_make_response(sk, dst, req);
289 if (skb != NULL) {
290 struct dccp_hdr *dh = dccp_hdr(skb);
292 dh->dccph_checksum = dccp_v6_csum_finish(skb,
293 &ireq6->loc_addr,
294 &ireq6->rmt_addr);
295 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
296 err = ip6_xmit(sk, skb, &fl, opt, 0);
297 err = net_xmit_eval(err);
300 done:
301 if (opt != NULL && opt != np->opt)
302 sock_kfree_s(sk, opt, opt->tot_len);
303 dst_release(dst);
304 return err;
307 static void dccp_v6_reqsk_destructor(struct request_sock *req)
309 if (inet6_rsk(req)->pktopts != NULL)
310 kfree_skb(inet6_rsk(req)->pktopts);
313 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
315 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
316 const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
317 sizeof(struct dccp_hdr_ext) +
318 sizeof(struct dccp_hdr_reset);
319 struct sk_buff *skb;
320 struct flowi fl;
321 u64 seqno = 0;
323 if (rxdh->dccph_type == DCCP_PKT_RESET)
324 return;
326 if (!ipv6_unicast_destination(rxskb))
327 return;
329 skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
330 GFP_ATOMIC);
331 if (skb == NULL)
332 return;
334 skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
336 dh = dccp_zeroed_hdr(skb, dccp_hdr_reset_len);
338 /* Swap the send and the receive. */
339 dh->dccph_type = DCCP_PKT_RESET;
340 dh->dccph_sport = rxdh->dccph_dport;
341 dh->dccph_dport = rxdh->dccph_sport;
342 dh->dccph_doff = dccp_hdr_reset_len / 4;
343 dh->dccph_x = 1;
344 dccp_hdr_reset(skb)->dccph_reset_code =
345 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
347 /* See "8.3.1. Abnormal Termination" in RFC 4340 */
348 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
349 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
351 dccp_hdr_set_seq(dh, seqno);
352 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), DCCP_SKB_CB(rxskb)->dccpd_seq);
354 dccp_csum_outgoing(skb);
355 dh->dccph_checksum = dccp_v6_csum_finish(skb, &rxskb->nh.ipv6h->saddr,
356 &rxskb->nh.ipv6h->daddr);
358 memset(&fl, 0, sizeof(fl));
359 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
360 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
362 fl.proto = IPPROTO_DCCP;
363 fl.oif = inet6_iif(rxskb);
364 fl.fl_ip_dport = dh->dccph_dport;
365 fl.fl_ip_sport = dh->dccph_sport;
366 security_skb_classify_flow(rxskb, &fl);
368 /* sk = NULL, but it is safe for now. RST socket required. */
369 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
370 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
371 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
372 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
373 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
374 return;
378 kfree_skb(skb);
381 static struct request_sock_ops dccp6_request_sock_ops = {
382 .family = AF_INET6,
383 .obj_size = sizeof(struct dccp6_request_sock),
384 .rtx_syn_ack = dccp_v6_send_response,
385 .send_ack = dccp_reqsk_send_ack,
386 .destructor = dccp_v6_reqsk_destructor,
387 .send_reset = dccp_v6_ctl_send_reset,
390 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
392 const struct dccp_hdr *dh = dccp_hdr(skb);
393 const struct ipv6hdr *iph = skb->nh.ipv6h;
394 struct sock *nsk;
395 struct request_sock **prev;
396 /* Find possible connection requests. */
397 struct request_sock *req = inet6_csk_search_req(sk, &prev,
398 dh->dccph_sport,
399 &iph->saddr,
400 &iph->daddr,
401 inet6_iif(skb));
402 if (req != NULL)
403 return dccp_check_req(sk, skb, req, prev);
405 nsk = __inet6_lookup_established(&dccp_hashinfo,
406 &iph->saddr, dh->dccph_sport,
407 &iph->daddr, ntohs(dh->dccph_dport),
408 inet6_iif(skb));
409 if (nsk != NULL) {
410 if (nsk->sk_state != DCCP_TIME_WAIT) {
411 bh_lock_sock(nsk);
412 return nsk;
414 inet_twsk_put(inet_twsk(nsk));
415 return NULL;
418 return sk;
421 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
423 struct request_sock *req;
424 struct dccp_request_sock *dreq;
425 struct inet6_request_sock *ireq6;
426 struct ipv6_pinfo *np = inet6_sk(sk);
427 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
428 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
429 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
431 if (skb->protocol == htons(ETH_P_IP))
432 return dccp_v4_conn_request(sk, skb);
434 if (!ipv6_unicast_destination(skb))
435 goto drop;
437 if (dccp_bad_service_code(sk, service)) {
438 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
439 goto drop;
442 * There are no SYN attacks on IPv6, yet...
444 if (inet_csk_reqsk_queue_is_full(sk))
445 goto drop;
447 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
448 goto drop;
450 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
451 if (req == NULL)
452 goto drop;
454 if (dccp_parse_options(sk, skb))
455 goto drop_and_free;
457 dccp_reqsk_init(req, skb);
459 if (security_inet_conn_request(sk, skb, req))
460 goto drop_and_free;
462 ireq6 = inet6_rsk(req);
463 ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
464 ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
465 ireq6->pktopts = NULL;
467 if (ipv6_opt_accepted(sk, skb) ||
468 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
469 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
470 atomic_inc(&skb->users);
471 ireq6->pktopts = skb;
473 ireq6->iif = sk->sk_bound_dev_if;
475 /* So that link locals have meaning */
476 if (!sk->sk_bound_dev_if &&
477 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
478 ireq6->iif = inet6_iif(skb);
481 * Step 3: Process LISTEN state
483 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
485 * In fact we defer setting S.GSR, S.SWL, S.SWH to
486 * dccp_create_openreq_child.
488 dreq = dccp_rsk(req);
489 dreq->dreq_isr = dcb->dccpd_seq;
490 dreq->dreq_iss = dccp_v6_init_sequence(skb);
491 dreq->dreq_service = service;
493 if (dccp_v6_send_response(sk, req, NULL))
494 goto drop_and_free;
496 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
497 return 0;
499 drop_and_free:
500 reqsk_free(req);
501 drop:
502 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
503 dcb->dccpd_reset_code = reset_code;
504 return -1;
507 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
508 struct sk_buff *skb,
509 struct request_sock *req,
510 struct dst_entry *dst)
512 struct inet6_request_sock *ireq6 = inet6_rsk(req);
513 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
514 struct inet_sock *newinet;
515 struct dccp_sock *newdp;
516 struct dccp6_sock *newdp6;
517 struct sock *newsk;
518 struct ipv6_txoptions *opt;
520 if (skb->protocol == htons(ETH_P_IP)) {
522 * v6 mapped
524 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
525 if (newsk == NULL)
526 return NULL;
528 newdp6 = (struct dccp6_sock *)newsk;
529 newdp = dccp_sk(newsk);
530 newinet = inet_sk(newsk);
531 newinet->pinet6 = &newdp6->inet6;
532 newnp = inet6_sk(newsk);
534 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
536 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
537 newinet->daddr);
539 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
540 newinet->saddr);
542 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
544 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
545 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
546 newnp->pktoptions = NULL;
547 newnp->opt = NULL;
548 newnp->mcast_oif = inet6_iif(skb);
549 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
552 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
553 * here, dccp_create_openreq_child now does this for us, see the comment in
554 * that function for the gory details. -acme
557 /* It is tricky place. Until this moment IPv4 tcp
558 worked with IPv6 icsk.icsk_af_ops.
559 Sync it now.
561 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
563 return newsk;
566 opt = np->opt;
568 if (sk_acceptq_is_full(sk))
569 goto out_overflow;
571 if (np->rxopt.bits.osrcrt == 2 && opt == NULL && ireq6->pktopts) {
572 const struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
574 if (rxopt->srcrt)
575 opt = ipv6_invert_rthdr(sk,
576 (struct ipv6_rt_hdr *)(ireq6->pktopts->nh.raw +
577 rxopt->srcrt));
580 if (dst == NULL) {
581 struct in6_addr *final_p = NULL, final;
582 struct flowi fl;
584 memset(&fl, 0, sizeof(fl));
585 fl.proto = IPPROTO_DCCP;
586 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
587 if (opt != NULL && opt->srcrt != NULL) {
588 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
590 ipv6_addr_copy(&final, &fl.fl6_dst);
591 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
592 final_p = &final;
594 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
595 fl.oif = sk->sk_bound_dev_if;
596 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
597 fl.fl_ip_sport = inet_sk(sk)->sport;
598 security_sk_classify_flow(sk, &fl);
600 if (ip6_dst_lookup(sk, &dst, &fl))
601 goto out;
603 if (final_p)
604 ipv6_addr_copy(&fl.fl6_dst, final_p);
606 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
607 goto out;
610 newsk = dccp_create_openreq_child(sk, req, skb);
611 if (newsk == NULL)
612 goto out;
615 * No need to charge this sock to the relevant IPv6 refcnt debug socks
616 * count here, dccp_create_openreq_child now does this for us, see the
617 * comment in that function for the gory details. -acme
620 __ip6_dst_store(newsk, dst, NULL, NULL);
621 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
622 NETIF_F_TSO);
623 newdp6 = (struct dccp6_sock *)newsk;
624 newinet = inet_sk(newsk);
625 newinet->pinet6 = &newdp6->inet6;
626 newdp = dccp_sk(newsk);
627 newnp = inet6_sk(newsk);
629 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
631 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
632 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
633 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
634 newsk->sk_bound_dev_if = ireq6->iif;
636 /* Now IPv6 options...
638 First: no IPv4 options.
640 newinet->opt = NULL;
642 /* Clone RX bits */
643 newnp->rxopt.all = np->rxopt.all;
645 /* Clone pktoptions received with SYN */
646 newnp->pktoptions = NULL;
647 if (ireq6->pktopts != NULL) {
648 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
649 kfree_skb(ireq6->pktopts);
650 ireq6->pktopts = NULL;
651 if (newnp->pktoptions)
652 skb_set_owner_r(newnp->pktoptions, newsk);
654 newnp->opt = NULL;
655 newnp->mcast_oif = inet6_iif(skb);
656 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
659 * Clone native IPv6 options from listening socket (if any)
661 * Yes, keeping reference count would be much more clever, but we make
662 * one more one thing there: reattach optmem to newsk.
664 if (opt != NULL) {
665 newnp->opt = ipv6_dup_options(newsk, opt);
666 if (opt != np->opt)
667 sock_kfree_s(sk, opt, opt->tot_len);
670 inet_csk(newsk)->icsk_ext_hdr_len = 0;
671 if (newnp->opt != NULL)
672 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
673 newnp->opt->opt_flen);
675 dccp_sync_mss(newsk, dst_mtu(dst));
677 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
679 __inet6_hash(&dccp_hashinfo, newsk);
680 inet_inherit_port(&dccp_hashinfo, sk, newsk);
682 return newsk;
684 out_overflow:
685 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
686 out:
687 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
688 if (opt != NULL && opt != np->opt)
689 sock_kfree_s(sk, opt, opt->tot_len);
690 dst_release(dst);
691 return NULL;
694 /* The socket must have it's spinlock held when we get
695 * here.
697 * We have a potential double-lock case here, so even when
698 * doing backlog processing we use the BH locking scheme.
699 * This is because we cannot sleep with the original spinlock
700 * held.
702 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
704 struct ipv6_pinfo *np = inet6_sk(sk);
705 struct sk_buff *opt_skb = NULL;
707 /* Imagine: socket is IPv6. IPv4 packet arrives,
708 goes to IPv4 receive handler and backlogged.
709 From backlog it always goes here. Kerboom...
710 Fortunately, dccp_rcv_established and rcv_established
711 handle them correctly, but it is not case with
712 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
715 if (skb->protocol == htons(ETH_P_IP))
716 return dccp_v4_do_rcv(sk, skb);
718 if (sk_filter(sk, skb))
719 goto discard;
722 * socket locking is here for SMP purposes as backlog rcv is currently
723 * called with bh processing disabled.
726 /* Do Stevens' IPV6_PKTOPTIONS.
728 Yes, guys, it is the only place in our code, where we
729 may make it not affecting IPv4.
730 The rest of code is protocol independent,
731 and I do not like idea to uglify IPv4.
733 Actually, all the idea behind IPV6_PKTOPTIONS
734 looks not very well thought. For now we latch
735 options, received in the last packet, enqueued
736 by tcp. Feel free to propose better solution.
737 --ANK (980728)
739 if (np->rxopt.all)
741 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
742 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
744 opt_skb = skb_clone(skb, GFP_ATOMIC);
746 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
747 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
748 goto reset;
749 if (opt_skb) {
750 /* XXX This is where we would goto ipv6_pktoptions. */
751 __kfree_skb(opt_skb);
753 return 0;
757 * Step 3: Process LISTEN state
758 * If S.state == LISTEN,
759 * If P.type == Request or P contains a valid Init Cookie option,
760 * (* Must scan the packet's options to check for Init
761 * Cookies. Only Init Cookies are processed here,
762 * however; other options are processed in Step 8. This
763 * scan need only be performed if the endpoint uses Init
764 * Cookies *)
765 * (* Generate a new socket and switch to that socket *)
766 * Set S := new socket for this port pair
767 * S.state = RESPOND
768 * Choose S.ISS (initial seqno) or set from Init Cookies
769 * Initialize S.GAR := S.ISS
770 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
771 * Continue with S.state == RESPOND
772 * (* A Response packet will be generated in Step 11 *)
773 * Otherwise,
774 * Generate Reset(No Connection) unless P.type == Reset
775 * Drop packet and return
777 * NOTE: the check for the packet types is done in
778 * dccp_rcv_state_process
780 if (sk->sk_state == DCCP_LISTEN) {
781 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
783 if (nsk == NULL)
784 goto discard;
786 * Queue it on the new socket if the new socket is active,
787 * otherwise we just shortcircuit this and continue with
788 * the new socket..
790 if (nsk != sk) {
791 if (dccp_child_process(sk, nsk, skb))
792 goto reset;
793 if (opt_skb != NULL)
794 __kfree_skb(opt_skb);
795 return 0;
799 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
800 goto reset;
801 if (opt_skb) {
802 /* XXX This is where we would goto ipv6_pktoptions. */
803 __kfree_skb(opt_skb);
805 return 0;
807 reset:
808 dccp_v6_ctl_send_reset(sk, skb);
809 discard:
810 if (opt_skb != NULL)
811 __kfree_skb(opt_skb);
812 kfree_skb(skb);
813 return 0;
816 static int dccp_v6_rcv(struct sk_buff **pskb)
818 const struct dccp_hdr *dh;
819 struct sk_buff *skb = *pskb;
820 struct sock *sk;
821 int min_cov;
823 /* Step 1: Check header basics */
825 if (dccp_invalid_packet(skb))
826 goto discard_it;
828 /* Step 1: If header checksum is incorrect, drop packet and return. */
829 if (dccp_v6_csum_finish(skb, &skb->nh.ipv6h->saddr,
830 &skb->nh.ipv6h->daddr)) {
831 DCCP_WARN("dropped packet with invalid checksum\n");
832 goto discard_it;
835 dh = dccp_hdr(skb);
837 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
838 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
840 if (dccp_packet_without_ack(skb))
841 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
842 else
843 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
845 /* Step 2:
846 * Look up flow ID in table and get corresponding socket */
847 sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
848 dh->dccph_sport,
849 &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
850 inet6_iif(skb));
852 * Step 2:
853 * If no socket ...
855 if (sk == NULL) {
856 dccp_pr_debug("failed to look up flow ID in table and "
857 "get corresponding socket\n");
858 goto no_dccp_socket;
862 * Step 2:
863 * ... or S.state == TIMEWAIT,
864 * Generate Reset(No Connection) unless P.type == Reset
865 * Drop packet and return
867 if (sk->sk_state == DCCP_TIME_WAIT) {
868 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
869 inet_twsk_put(inet_twsk(sk));
870 goto no_dccp_socket;
874 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
875 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
876 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
878 min_cov = dccp_sk(sk)->dccps_pcrlen;
879 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
880 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
881 dh->dccph_cscov, min_cov);
882 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
883 goto discard_and_relse;
886 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
887 goto discard_and_relse;
889 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
891 no_dccp_socket:
892 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
893 goto discard_it;
895 * Step 2:
896 * If no socket ...
897 * Generate Reset(No Connection) unless P.type == Reset
898 * Drop packet and return
900 if (dh->dccph_type != DCCP_PKT_RESET) {
901 DCCP_SKB_CB(skb)->dccpd_reset_code =
902 DCCP_RESET_CODE_NO_CONNECTION;
903 dccp_v6_ctl_send_reset(sk, skb);
906 discard_it:
907 kfree_skb(skb);
908 return 0;
910 discard_and_relse:
911 sock_put(sk);
912 goto discard_it;
915 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
916 int addr_len)
918 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
919 struct inet_connection_sock *icsk = inet_csk(sk);
920 struct inet_sock *inet = inet_sk(sk);
921 struct ipv6_pinfo *np = inet6_sk(sk);
922 struct dccp_sock *dp = dccp_sk(sk);
923 struct in6_addr *saddr = NULL, *final_p = NULL, final;
924 struct flowi fl;
925 struct dst_entry *dst;
926 int addr_type;
927 int err;
929 dp->dccps_role = DCCP_ROLE_CLIENT;
931 if (addr_len < SIN6_LEN_RFC2133)
932 return -EINVAL;
934 if (usin->sin6_family != AF_INET6)
935 return -EAFNOSUPPORT;
937 memset(&fl, 0, sizeof(fl));
939 if (np->sndflow) {
940 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
941 IP6_ECN_flow_init(fl.fl6_flowlabel);
942 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
943 struct ip6_flowlabel *flowlabel;
944 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
945 if (flowlabel == NULL)
946 return -EINVAL;
947 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
948 fl6_sock_release(flowlabel);
952 * connect() to INADDR_ANY means loopback (BSD'ism).
954 if (ipv6_addr_any(&usin->sin6_addr))
955 usin->sin6_addr.s6_addr[15] = 1;
957 addr_type = ipv6_addr_type(&usin->sin6_addr);
959 if (addr_type & IPV6_ADDR_MULTICAST)
960 return -ENETUNREACH;
962 if (addr_type & IPV6_ADDR_LINKLOCAL) {
963 if (addr_len >= sizeof(struct sockaddr_in6) &&
964 usin->sin6_scope_id) {
965 /* If interface is set while binding, indices
966 * must coincide.
968 if (sk->sk_bound_dev_if &&
969 sk->sk_bound_dev_if != usin->sin6_scope_id)
970 return -EINVAL;
972 sk->sk_bound_dev_if = usin->sin6_scope_id;
975 /* Connect to link-local address requires an interface */
976 if (!sk->sk_bound_dev_if)
977 return -EINVAL;
980 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
981 np->flow_label = fl.fl6_flowlabel;
984 * DCCP over IPv4
986 if (addr_type == IPV6_ADDR_MAPPED) {
987 u32 exthdrlen = icsk->icsk_ext_hdr_len;
988 struct sockaddr_in sin;
990 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
992 if (__ipv6_only_sock(sk))
993 return -ENETUNREACH;
995 sin.sin_family = AF_INET;
996 sin.sin_port = usin->sin6_port;
997 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
999 icsk->icsk_af_ops = &dccp_ipv6_mapped;
1000 sk->sk_backlog_rcv = dccp_v4_do_rcv;
1002 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
1003 if (err) {
1004 icsk->icsk_ext_hdr_len = exthdrlen;
1005 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
1006 sk->sk_backlog_rcv = dccp_v6_do_rcv;
1007 goto failure;
1008 } else {
1009 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
1010 inet->saddr);
1011 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
1012 inet->rcv_saddr);
1015 return err;
1018 if (!ipv6_addr_any(&np->rcv_saddr))
1019 saddr = &np->rcv_saddr;
1021 fl.proto = IPPROTO_DCCP;
1022 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
1023 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
1024 fl.oif = sk->sk_bound_dev_if;
1025 fl.fl_ip_dport = usin->sin6_port;
1026 fl.fl_ip_sport = inet->sport;
1027 security_sk_classify_flow(sk, &fl);
1029 if (np->opt != NULL && np->opt->srcrt != NULL) {
1030 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
1032 ipv6_addr_copy(&final, &fl.fl6_dst);
1033 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1034 final_p = &final;
1037 err = ip6_dst_lookup(sk, &dst, &fl);
1038 if (err)
1039 goto failure;
1041 if (final_p)
1042 ipv6_addr_copy(&fl.fl6_dst, final_p);
1044 err = xfrm_lookup(&dst, &fl, sk, 0);
1045 if (err < 0)
1046 goto failure;
1048 if (saddr == NULL) {
1049 saddr = &fl.fl6_src;
1050 ipv6_addr_copy(&np->rcv_saddr, saddr);
1053 /* set the source address */
1054 ipv6_addr_copy(&np->saddr, saddr);
1055 inet->rcv_saddr = LOOPBACK4_IPV6;
1057 __ip6_dst_store(sk, dst, NULL, NULL);
1059 icsk->icsk_ext_hdr_len = 0;
1060 if (np->opt != NULL)
1061 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1062 np->opt->opt_nflen);
1064 inet->dport = usin->sin6_port;
1066 dccp_set_state(sk, DCCP_REQUESTING);
1067 err = inet6_hash_connect(&dccp_death_row, sk);
1068 if (err)
1069 goto late_failure;
1071 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1072 np->daddr.s6_addr32,
1073 inet->sport, inet->dport);
1074 err = dccp_connect(sk);
1075 if (err)
1076 goto late_failure;
1078 return 0;
1080 late_failure:
1081 dccp_set_state(sk, DCCP_CLOSED);
1082 __sk_dst_reset(sk);
1083 failure:
1084 inet->dport = 0;
1085 sk->sk_route_caps = 0;
1086 return err;
1089 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1090 .queue_xmit = inet6_csk_xmit,
1091 .send_check = dccp_v6_send_check,
1092 .rebuild_header = inet6_sk_rebuild_header,
1093 .conn_request = dccp_v6_conn_request,
1094 .syn_recv_sock = dccp_v6_request_recv_sock,
1095 .net_header_len = sizeof(struct ipv6hdr),
1096 .setsockopt = ipv6_setsockopt,
1097 .getsockopt = ipv6_getsockopt,
1098 .addr2sockaddr = inet6_csk_addr2sockaddr,
1099 .sockaddr_len = sizeof(struct sockaddr_in6),
1100 #ifdef CONFIG_COMPAT
1101 .compat_setsockopt = compat_ipv6_setsockopt,
1102 .compat_getsockopt = compat_ipv6_getsockopt,
1103 #endif
1107 * DCCP over IPv4 via INET6 API
1109 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1110 .queue_xmit = ip_queue_xmit,
1111 .send_check = dccp_v4_send_check,
1112 .rebuild_header = inet_sk_rebuild_header,
1113 .conn_request = dccp_v6_conn_request,
1114 .syn_recv_sock = dccp_v6_request_recv_sock,
1115 .net_header_len = sizeof(struct iphdr),
1116 .setsockopt = ipv6_setsockopt,
1117 .getsockopt = ipv6_getsockopt,
1118 .addr2sockaddr = inet6_csk_addr2sockaddr,
1119 .sockaddr_len = sizeof(struct sockaddr_in6),
1120 #ifdef CONFIG_COMPAT
1121 .compat_setsockopt = compat_ipv6_setsockopt,
1122 .compat_getsockopt = compat_ipv6_getsockopt,
1123 #endif
1126 /* NOTE: A lot of things set to zero explicitly by call to
1127 * sk_alloc() so need not be done here.
1129 static int dccp_v6_init_sock(struct sock *sk)
1131 static __u8 dccp_v6_ctl_sock_initialized;
1132 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1134 if (err == 0) {
1135 if (unlikely(!dccp_v6_ctl_sock_initialized))
1136 dccp_v6_ctl_sock_initialized = 1;
1137 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1140 return err;
1143 static int dccp_v6_destroy_sock(struct sock *sk)
1145 dccp_destroy_sock(sk);
1146 return inet6_destroy_sock(sk);
1149 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1150 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1153 static struct proto dccp_v6_prot = {
1154 .name = "DCCPv6",
1155 .owner = THIS_MODULE,
1156 .close = dccp_close,
1157 .connect = dccp_v6_connect,
1158 .disconnect = dccp_disconnect,
1159 .ioctl = dccp_ioctl,
1160 .init = dccp_v6_init_sock,
1161 .setsockopt = dccp_setsockopt,
1162 .getsockopt = dccp_getsockopt,
1163 .sendmsg = dccp_sendmsg,
1164 .recvmsg = dccp_recvmsg,
1165 .backlog_rcv = dccp_v6_do_rcv,
1166 .hash = dccp_v6_hash,
1167 .unhash = dccp_unhash,
1168 .accept = inet_csk_accept,
1169 .get_port = dccp_v6_get_port,
1170 .shutdown = dccp_shutdown,
1171 .destroy = dccp_v6_destroy_sock,
1172 .orphan_count = &dccp_orphan_count,
1173 .max_header = MAX_DCCP_HEADER,
1174 .obj_size = sizeof(struct dccp6_sock),
1175 .rsk_prot = &dccp6_request_sock_ops,
1176 .twsk_prot = &dccp6_timewait_sock_ops,
1177 #ifdef CONFIG_COMPAT
1178 .compat_setsockopt = compat_dccp_setsockopt,
1179 .compat_getsockopt = compat_dccp_getsockopt,
1180 #endif
1183 static struct inet6_protocol dccp_v6_protocol = {
1184 .handler = dccp_v6_rcv,
1185 .err_handler = dccp_v6_err,
1186 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1189 static struct proto_ops inet6_dccp_ops = {
1190 .family = PF_INET6,
1191 .owner = THIS_MODULE,
1192 .release = inet6_release,
1193 .bind = inet6_bind,
1194 .connect = inet_stream_connect,
1195 .socketpair = sock_no_socketpair,
1196 .accept = inet_accept,
1197 .getname = inet6_getname,
1198 .poll = dccp_poll,
1199 .ioctl = inet6_ioctl,
1200 .listen = inet_dccp_listen,
1201 .shutdown = inet_shutdown,
1202 .setsockopt = sock_common_setsockopt,
1203 .getsockopt = sock_common_getsockopt,
1204 .sendmsg = inet_sendmsg,
1205 .recvmsg = sock_common_recvmsg,
1206 .mmap = sock_no_mmap,
1207 .sendpage = sock_no_sendpage,
1208 #ifdef CONFIG_COMPAT
1209 .compat_setsockopt = compat_sock_common_setsockopt,
1210 .compat_getsockopt = compat_sock_common_getsockopt,
1211 #endif
1214 static struct inet_protosw dccp_v6_protosw = {
1215 .type = SOCK_DCCP,
1216 .protocol = IPPROTO_DCCP,
1217 .prot = &dccp_v6_prot,
1218 .ops = &inet6_dccp_ops,
1219 .capability = -1,
1220 .flags = INET_PROTOSW_ICSK,
1223 static int __init dccp_v6_init(void)
1225 int err = proto_register(&dccp_v6_prot, 1);
1227 if (err != 0)
1228 goto out;
1230 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1231 if (err != 0)
1232 goto out_unregister_proto;
1234 inet6_register_protosw(&dccp_v6_protosw);
1236 err = inet_csk_ctl_sock_create(&dccp_v6_ctl_socket, PF_INET6,
1237 SOCK_DCCP, IPPROTO_DCCP);
1238 if (err != 0)
1239 goto out_unregister_protosw;
1240 out:
1241 return err;
1242 out_unregister_protosw:
1243 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1244 inet6_unregister_protosw(&dccp_v6_protosw);
1245 out_unregister_proto:
1246 proto_unregister(&dccp_v6_prot);
1247 goto out;
1250 static void __exit dccp_v6_exit(void)
1252 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1253 inet6_unregister_protosw(&dccp_v6_protosw);
1254 proto_unregister(&dccp_v6_prot);
1257 module_init(dccp_v6_init);
1258 module_exit(dccp_v6_exit);
1261 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1262 * values directly, Also cover the case where the protocol is not specified,
1263 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1265 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
1266 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
1267 MODULE_LICENSE("GPL");
1268 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1269 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");