Linux v2.6.17-rc2
[linux-2.6/next.git] / net / dccp / ipv6.c
blob65e2ab0886e62fd3a2bddde3e9d961b40b63a195
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/config.h>
16 #include <linux/module.h>
17 #include <linux/random.h>
18 #include <linux/xfrm.h>
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>
33 #include "dccp.h"
34 #include "ipv6.h"
36 /* Socket used for sending RSTs and ACKs */
37 static struct socket *dccp_v6_ctl_socket;
39 static void dccp_v6_ctl_send_reset(struct sk_buff *skb);
40 static void dccp_v6_reqsk_send_ack(struct sk_buff *skb,
41 struct request_sock *req);
42 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb);
44 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
46 static struct inet_connection_sock_af_ops dccp_ipv6_mapped;
47 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
49 static int dccp_v6_get_port(struct sock *sk, unsigned short snum)
51 return inet_csk_get_port(&dccp_hashinfo, sk, snum,
52 inet6_csk_bind_conflict);
55 static void dccp_v6_hash(struct sock *sk)
57 if (sk->sk_state != DCCP_CLOSED) {
58 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
59 dccp_hash(sk);
60 return;
62 local_bh_disable();
63 __inet6_hash(&dccp_hashinfo, sk);
64 local_bh_enable();
68 static inline u16 dccp_v6_check(struct dccp_hdr *dh, int len,
69 struct in6_addr *saddr,
70 struct in6_addr *daddr,
71 unsigned long base)
73 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_DCCP, base);
76 static __u32 dccp_v6_init_sequence(struct sock *sk, struct sk_buff *skb)
78 const struct dccp_hdr *dh = dccp_hdr(skb);
80 if (skb->protocol == htons(ETH_P_IPV6))
81 return secure_tcpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
82 skb->nh.ipv6h->saddr.s6_addr32,
83 dh->dccph_dport,
84 dh->dccph_sport);
86 return secure_dccp_sequence_number(skb->nh.iph->daddr,
87 skb->nh.iph->saddr,
88 dh->dccph_dport,
89 dh->dccph_sport);
92 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
93 int addr_len)
95 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
96 struct inet_connection_sock *icsk = inet_csk(sk);
97 struct inet_sock *inet = inet_sk(sk);
98 struct ipv6_pinfo *np = inet6_sk(sk);
99 struct dccp_sock *dp = dccp_sk(sk);
100 struct in6_addr *saddr = NULL, *final_p = NULL, final;
101 struct flowi fl;
102 struct dst_entry *dst;
103 int addr_type;
104 int err;
106 dp->dccps_role = DCCP_ROLE_CLIENT;
108 if (addr_len < SIN6_LEN_RFC2133)
109 return -EINVAL;
111 if (usin->sin6_family != AF_INET6)
112 return -EAFNOSUPPORT;
114 memset(&fl, 0, sizeof(fl));
116 if (np->sndflow) {
117 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
118 IP6_ECN_flow_init(fl.fl6_flowlabel);
119 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
120 struct ip6_flowlabel *flowlabel;
121 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
122 if (flowlabel == NULL)
123 return -EINVAL;
124 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
125 fl6_sock_release(flowlabel);
129 * connect() to INADDR_ANY means loopback (BSD'ism).
131 if (ipv6_addr_any(&usin->sin6_addr))
132 usin->sin6_addr.s6_addr[15] = 1;
134 addr_type = ipv6_addr_type(&usin->sin6_addr);
136 if (addr_type & IPV6_ADDR_MULTICAST)
137 return -ENETUNREACH;
139 if (addr_type & IPV6_ADDR_LINKLOCAL) {
140 if (addr_len >= sizeof(struct sockaddr_in6) &&
141 usin->sin6_scope_id) {
142 /* If interface is set while binding, indices
143 * must coincide.
145 if (sk->sk_bound_dev_if &&
146 sk->sk_bound_dev_if != usin->sin6_scope_id)
147 return -EINVAL;
149 sk->sk_bound_dev_if = usin->sin6_scope_id;
152 /* Connect to link-local address requires an interface */
153 if (!sk->sk_bound_dev_if)
154 return -EINVAL;
157 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
158 np->flow_label = fl.fl6_flowlabel;
161 * DCCP over IPv4
163 if (addr_type == IPV6_ADDR_MAPPED) {
164 u32 exthdrlen = icsk->icsk_ext_hdr_len;
165 struct sockaddr_in sin;
167 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
169 if (__ipv6_only_sock(sk))
170 return -ENETUNREACH;
172 sin.sin_family = AF_INET;
173 sin.sin_port = usin->sin6_port;
174 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
176 icsk->icsk_af_ops = &dccp_ipv6_mapped;
177 sk->sk_backlog_rcv = dccp_v4_do_rcv;
179 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
180 if (err) {
181 icsk->icsk_ext_hdr_len = exthdrlen;
182 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
183 sk->sk_backlog_rcv = dccp_v6_do_rcv;
184 goto failure;
185 } else {
186 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
187 inet->saddr);
188 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
189 inet->rcv_saddr);
192 return err;
195 if (!ipv6_addr_any(&np->rcv_saddr))
196 saddr = &np->rcv_saddr;
198 fl.proto = IPPROTO_DCCP;
199 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
200 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
201 fl.oif = sk->sk_bound_dev_if;
202 fl.fl_ip_dport = usin->sin6_port;
203 fl.fl_ip_sport = inet->sport;
205 if (np->opt != NULL && np->opt->srcrt != NULL) {
206 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
208 ipv6_addr_copy(&final, &fl.fl6_dst);
209 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
210 final_p = &final;
213 err = ip6_dst_lookup(sk, &dst, &fl);
214 if (err)
215 goto failure;
217 if (final_p)
218 ipv6_addr_copy(&fl.fl6_dst, final_p);
220 err = xfrm_lookup(&dst, &fl, sk, 0);
221 if (err < 0)
222 goto failure;
224 if (saddr == NULL) {
225 saddr = &fl.fl6_src;
226 ipv6_addr_copy(&np->rcv_saddr, saddr);
229 /* set the source address */
230 ipv6_addr_copy(&np->saddr, saddr);
231 inet->rcv_saddr = LOOPBACK4_IPV6;
233 ip6_dst_store(sk, dst, NULL);
235 icsk->icsk_ext_hdr_len = 0;
236 if (np->opt != NULL)
237 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
238 np->opt->opt_nflen);
240 inet->dport = usin->sin6_port;
242 dccp_set_state(sk, DCCP_REQUESTING);
243 err = inet6_hash_connect(&dccp_death_row, sk);
244 if (err)
245 goto late_failure;
246 /* FIXME */
247 #if 0
248 dp->dccps_gar = secure_dccp_v6_sequence_number(np->saddr.s6_addr32,
249 np->daddr.s6_addr32,
250 inet->sport,
251 inet->dport);
252 #endif
253 err = dccp_connect(sk);
254 if (err)
255 goto late_failure;
257 return 0;
259 late_failure:
260 dccp_set_state(sk, DCCP_CLOSED);
261 __sk_dst_reset(sk);
262 failure:
263 inet->dport = 0;
264 sk->sk_route_caps = 0;
265 return err;
268 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
269 int type, int code, int offset, __be32 info)
271 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
272 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
273 struct ipv6_pinfo *np;
274 struct sock *sk;
275 int err;
276 __u64 seq;
278 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
279 &hdr->saddr, dh->dccph_sport, skb->dev->ifindex);
281 if (sk == NULL) {
282 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
283 return;
286 if (sk->sk_state == DCCP_TIME_WAIT) {
287 inet_twsk_put((struct inet_timewait_sock *)sk);
288 return;
291 bh_lock_sock(sk);
292 if (sock_owned_by_user(sk))
293 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
295 if (sk->sk_state == DCCP_CLOSED)
296 goto out;
298 np = inet6_sk(sk);
300 if (type == ICMPV6_PKT_TOOBIG) {
301 struct dst_entry *dst = NULL;
303 if (sock_owned_by_user(sk))
304 goto out;
305 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
306 goto out;
308 /* icmp should have updated the destination cache entry */
309 dst = __sk_dst_check(sk, np->dst_cookie);
310 if (dst == NULL) {
311 struct inet_sock *inet = inet_sk(sk);
312 struct flowi fl;
314 /* BUGGG_FUTURE: Again, it is not clear how
315 to handle rthdr case. Ignore this complexity
316 for now.
318 memset(&fl, 0, sizeof(fl));
319 fl.proto = IPPROTO_DCCP;
320 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
321 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
322 fl.oif = sk->sk_bound_dev_if;
323 fl.fl_ip_dport = inet->dport;
324 fl.fl_ip_sport = inet->sport;
326 err = ip6_dst_lookup(sk, &dst, &fl);
327 if (err) {
328 sk->sk_err_soft = -err;
329 goto out;
332 err = xfrm_lookup(&dst, &fl, sk, 0);
333 if (err < 0) {
334 sk->sk_err_soft = -err;
335 goto out;
337 } else
338 dst_hold(dst);
340 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
341 dccp_sync_mss(sk, dst_mtu(dst));
342 } /* else let the usual retransmit timer handle it */
343 dst_release(dst);
344 goto out;
347 icmpv6_err_convert(type, code, &err);
349 seq = DCCP_SKB_CB(skb)->dccpd_seq;
350 /* Might be for an request_sock */
351 switch (sk->sk_state) {
352 struct request_sock *req, **prev;
353 case DCCP_LISTEN:
354 if (sock_owned_by_user(sk))
355 goto out;
357 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
358 &hdr->daddr, &hdr->saddr,
359 inet6_iif(skb));
360 if (req == NULL)
361 goto out;
364 * ICMPs are not backlogged, hence we cannot get an established
365 * socket here.
367 BUG_TRAP(req->sk == NULL);
369 if (seq != dccp_rsk(req)->dreq_iss) {
370 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
371 goto out;
374 inet_csk_reqsk_queue_drop(sk, req, prev);
375 goto out;
377 case DCCP_REQUESTING:
378 case DCCP_RESPOND: /* Cannot happen.
379 It can, it SYNs are crossed. --ANK */
380 if (!sock_owned_by_user(sk)) {
381 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
382 sk->sk_err = err;
384 * Wake people up to see the error
385 * (see connect in sock.c)
387 sk->sk_error_report(sk);
388 dccp_done(sk);
389 } else
390 sk->sk_err_soft = err;
391 goto out;
394 if (!sock_owned_by_user(sk) && np->recverr) {
395 sk->sk_err = err;
396 sk->sk_error_report(sk);
397 } else
398 sk->sk_err_soft = err;
400 out:
401 bh_unlock_sock(sk);
402 sock_put(sk);
406 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
407 struct dst_entry *dst)
409 struct inet6_request_sock *ireq6 = inet6_rsk(req);
410 struct ipv6_pinfo *np = inet6_sk(sk);
411 struct sk_buff *skb;
412 struct ipv6_txoptions *opt = NULL;
413 struct in6_addr *final_p = NULL, final;
414 struct flowi fl;
415 int err = -1;
417 memset(&fl, 0, sizeof(fl));
418 fl.proto = IPPROTO_DCCP;
419 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
420 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
421 fl.fl6_flowlabel = 0;
422 fl.oif = ireq6->iif;
423 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
424 fl.fl_ip_sport = inet_sk(sk)->sport;
426 if (dst == NULL) {
427 opt = np->opt;
428 if (opt == NULL &&
429 np->rxopt.bits.osrcrt == 2 &&
430 ireq6->pktopts) {
431 struct sk_buff *pktopts = ireq6->pktopts;
432 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
434 if (rxopt->srcrt)
435 opt = ipv6_invert_rthdr(sk,
436 (struct ipv6_rt_hdr *)(pktopts->nh.raw +
437 rxopt->srcrt));
440 if (opt != NULL && opt->srcrt != NULL) {
441 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
443 ipv6_addr_copy(&final, &fl.fl6_dst);
444 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
445 final_p = &final;
448 err = ip6_dst_lookup(sk, &dst, &fl);
449 if (err)
450 goto done;
452 if (final_p)
453 ipv6_addr_copy(&fl.fl6_dst, final_p);
455 err = xfrm_lookup(&dst, &fl, sk, 0);
456 if (err < 0)
457 goto done;
460 skb = dccp_make_response(sk, dst, req);
461 if (skb != NULL) {
462 struct dccp_hdr *dh = dccp_hdr(skb);
464 dh->dccph_checksum = dccp_v6_check(dh, skb->len,
465 &ireq6->loc_addr,
466 &ireq6->rmt_addr,
467 csum_partial((char *)dh,
468 skb->len,
469 skb->csum));
470 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
471 err = ip6_xmit(sk, skb, &fl, opt, 0);
472 if (err == NET_XMIT_CN)
473 err = 0;
476 done:
477 if (opt != NULL && opt != np->opt)
478 sock_kfree_s(sk, opt, opt->tot_len);
479 dst_release(dst);
480 return err;
483 static void dccp_v6_reqsk_destructor(struct request_sock *req)
485 if (inet6_rsk(req)->pktopts != NULL)
486 kfree_skb(inet6_rsk(req)->pktopts);
489 static struct request_sock_ops dccp6_request_sock_ops = {
490 .family = AF_INET6,
491 .obj_size = sizeof(struct dccp6_request_sock),
492 .rtx_syn_ack = dccp_v6_send_response,
493 .send_ack = dccp_v6_reqsk_send_ack,
494 .destructor = dccp_v6_reqsk_destructor,
495 .send_reset = dccp_v6_ctl_send_reset,
498 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
499 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
502 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
504 struct ipv6_pinfo *np = inet6_sk(sk);
505 struct dccp_hdr *dh = dccp_hdr(skb);
507 dh->dccph_checksum = csum_ipv6_magic(&np->saddr, &np->daddr,
508 len, IPPROTO_DCCP,
509 csum_partial((char *)dh,
510 dh->dccph_doff << 2,
511 skb->csum));
514 static void dccp_v6_ctl_send_reset(struct sk_buff *rxskb)
516 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
517 const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
518 sizeof(struct dccp_hdr_ext) +
519 sizeof(struct dccp_hdr_reset);
520 struct sk_buff *skb;
521 struct flowi fl;
522 u64 seqno;
524 if (rxdh->dccph_type == DCCP_PKT_RESET)
525 return;
527 if (!ipv6_unicast_destination(rxskb))
528 return;
530 skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
531 GFP_ATOMIC);
532 if (skb == NULL)
533 return;
535 skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
537 skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
538 dh = dccp_hdr(skb);
539 memset(dh, 0, dccp_hdr_reset_len);
541 /* Swap the send and the receive. */
542 dh->dccph_type = DCCP_PKT_RESET;
543 dh->dccph_sport = rxdh->dccph_dport;
544 dh->dccph_dport = rxdh->dccph_sport;
545 dh->dccph_doff = dccp_hdr_reset_len / 4;
546 dh->dccph_x = 1;
547 dccp_hdr_reset(skb)->dccph_reset_code =
548 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
550 /* See "8.3.1. Abnormal Termination" in draft-ietf-dccp-spec-11 */
551 seqno = 0;
552 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
553 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
555 dccp_hdr_set_seq(dh, seqno);
556 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
557 DCCP_SKB_CB(rxskb)->dccpd_seq);
559 memset(&fl, 0, sizeof(fl));
560 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
561 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
562 dh->dccph_checksum = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
563 sizeof(*dh), IPPROTO_DCCP,
564 skb->csum);
565 fl.proto = IPPROTO_DCCP;
566 fl.oif = inet6_iif(rxskb);
567 fl.fl_ip_dport = dh->dccph_dport;
568 fl.fl_ip_sport = dh->dccph_sport;
570 /* sk = NULL, but it is safe for now. RST socket required. */
571 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
572 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
573 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
574 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
575 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
576 return;
580 kfree_skb(skb);
583 static void dccp_v6_reqsk_send_ack(struct sk_buff *rxskb,
584 struct request_sock *req)
586 struct flowi fl;
587 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
588 const u32 dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
589 sizeof(struct dccp_hdr_ext) +
590 sizeof(struct dccp_hdr_ack_bits);
591 struct sk_buff *skb;
593 skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
594 GFP_ATOMIC);
595 if (skb == NULL)
596 return;
598 skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
600 skb->h.raw = skb_push(skb, dccp_hdr_ack_len);
601 dh = dccp_hdr(skb);
602 memset(dh, 0, dccp_hdr_ack_len);
604 /* Build DCCP header and checksum it. */
605 dh->dccph_type = DCCP_PKT_ACK;
606 dh->dccph_sport = rxdh->dccph_dport;
607 dh->dccph_dport = rxdh->dccph_sport;
608 dh->dccph_doff = dccp_hdr_ack_len / 4;
609 dh->dccph_x = 1;
611 dccp_hdr_set_seq(dh, DCCP_SKB_CB(rxskb)->dccpd_ack_seq);
612 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
613 DCCP_SKB_CB(rxskb)->dccpd_seq);
615 memset(&fl, 0, sizeof(fl));
616 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
617 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
619 /* FIXME: calculate checksum, IPv4 also should... */
621 fl.proto = IPPROTO_DCCP;
622 fl.oif = inet6_iif(rxskb);
623 fl.fl_ip_dport = dh->dccph_dport;
624 fl.fl_ip_sport = dh->dccph_sport;
626 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
627 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
628 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
629 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
630 return;
634 kfree_skb(skb);
637 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
639 const struct dccp_hdr *dh = dccp_hdr(skb);
640 const struct ipv6hdr *iph = skb->nh.ipv6h;
641 struct sock *nsk;
642 struct request_sock **prev;
643 /* Find possible connection requests. */
644 struct request_sock *req = inet6_csk_search_req(sk, &prev,
645 dh->dccph_sport,
646 &iph->saddr,
647 &iph->daddr,
648 inet6_iif(skb));
649 if (req != NULL)
650 return dccp_check_req(sk, skb, req, prev);
652 nsk = __inet6_lookup_established(&dccp_hashinfo,
653 &iph->saddr, dh->dccph_sport,
654 &iph->daddr, ntohs(dh->dccph_dport),
655 inet6_iif(skb));
656 if (nsk != NULL) {
657 if (nsk->sk_state != DCCP_TIME_WAIT) {
658 bh_lock_sock(nsk);
659 return nsk;
661 inet_twsk_put((struct inet_timewait_sock *)nsk);
662 return NULL;
665 return sk;
668 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
670 struct inet_request_sock *ireq;
671 struct dccp_sock dp;
672 struct request_sock *req;
673 struct dccp_request_sock *dreq;
674 struct inet6_request_sock *ireq6;
675 struct ipv6_pinfo *np = inet6_sk(sk);
676 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
677 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
678 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
680 if (skb->protocol == htons(ETH_P_IP))
681 return dccp_v4_conn_request(sk, skb);
683 if (!ipv6_unicast_destination(skb))
684 goto drop;
686 if (dccp_bad_service_code(sk, service)) {
687 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
688 goto drop;
691 * There are no SYN attacks on IPv6, yet...
693 if (inet_csk_reqsk_queue_is_full(sk))
694 goto drop;
696 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
697 goto drop;
699 req = inet6_reqsk_alloc(sk->sk_prot->rsk_prot);
700 if (req == NULL)
701 goto drop;
703 /* FIXME: process options */
705 dccp_openreq_init(req, &dp, skb);
707 ireq6 = inet6_rsk(req);
708 ireq = inet_rsk(req);
709 ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
710 ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
711 req->rcv_wnd = 100; /* Fake, option parsing will get the
712 right value */
713 ireq6->pktopts = NULL;
715 if (ipv6_opt_accepted(sk, skb) ||
716 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
717 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
718 atomic_inc(&skb->users);
719 ireq6->pktopts = skb;
721 ireq6->iif = sk->sk_bound_dev_if;
723 /* So that link locals have meaning */
724 if (!sk->sk_bound_dev_if &&
725 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
726 ireq6->iif = inet6_iif(skb);
729 * Step 3: Process LISTEN state
731 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
733 * In fact we defer setting S.GSR, S.SWL, S.SWH to
734 * dccp_create_openreq_child.
736 dreq = dccp_rsk(req);
737 dreq->dreq_isr = dcb->dccpd_seq;
738 dreq->dreq_iss = dccp_v6_init_sequence(sk, skb);
739 dreq->dreq_service = service;
741 if (dccp_v6_send_response(sk, req, NULL))
742 goto drop_and_free;
744 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
745 return 0;
747 drop_and_free:
748 reqsk_free(req);
749 drop:
750 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
751 dcb->dccpd_reset_code = reset_code;
752 return -1;
755 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
756 struct sk_buff *skb,
757 struct request_sock *req,
758 struct dst_entry *dst)
760 struct inet6_request_sock *ireq6 = inet6_rsk(req);
761 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
762 struct inet_sock *newinet;
763 struct dccp_sock *newdp;
764 struct dccp6_sock *newdp6;
765 struct sock *newsk;
766 struct ipv6_txoptions *opt;
768 if (skb->protocol == htons(ETH_P_IP)) {
770 * v6 mapped
772 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
773 if (newsk == NULL)
774 return NULL;
776 newdp6 = (struct dccp6_sock *)newsk;
777 newdp = dccp_sk(newsk);
778 newinet = inet_sk(newsk);
779 newinet->pinet6 = &newdp6->inet6;
780 newnp = inet6_sk(newsk);
782 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
784 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
785 newinet->daddr);
787 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
788 newinet->saddr);
790 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
792 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
793 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
794 newnp->pktoptions = NULL;
795 newnp->opt = NULL;
796 newnp->mcast_oif = inet6_iif(skb);
797 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
800 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
801 * here, dccp_create_openreq_child now does this for us, see the comment in
802 * that function for the gory details. -acme
805 /* It is tricky place. Until this moment IPv4 tcp
806 worked with IPv6 icsk.icsk_af_ops.
807 Sync it now.
809 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
811 return newsk;
814 opt = np->opt;
816 if (sk_acceptq_is_full(sk))
817 goto out_overflow;
819 if (np->rxopt.bits.osrcrt == 2 && opt == NULL && ireq6->pktopts) {
820 const struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
822 if (rxopt->srcrt)
823 opt = ipv6_invert_rthdr(sk,
824 (struct ipv6_rt_hdr *)(ireq6->pktopts->nh.raw +
825 rxopt->srcrt));
828 if (dst == NULL) {
829 struct in6_addr *final_p = NULL, final;
830 struct flowi fl;
832 memset(&fl, 0, sizeof(fl));
833 fl.proto = IPPROTO_DCCP;
834 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
835 if (opt != NULL && opt->srcrt != NULL) {
836 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
838 ipv6_addr_copy(&final, &fl.fl6_dst);
839 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
840 final_p = &final;
842 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
843 fl.oif = sk->sk_bound_dev_if;
844 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
845 fl.fl_ip_sport = inet_sk(sk)->sport;
847 if (ip6_dst_lookup(sk, &dst, &fl))
848 goto out;
850 if (final_p)
851 ipv6_addr_copy(&fl.fl6_dst, final_p);
853 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
854 goto out;
857 newsk = dccp_create_openreq_child(sk, req, skb);
858 if (newsk == NULL)
859 goto out;
862 * No need to charge this sock to the relevant IPv6 refcnt debug socks
863 * count here, dccp_create_openreq_child now does this for us, see the
864 * comment in that function for the gory details. -acme
867 ip6_dst_store(newsk, dst, NULL);
868 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
869 NETIF_F_TSO);
870 newdp6 = (struct dccp6_sock *)newsk;
871 newinet = inet_sk(newsk);
872 newinet->pinet6 = &newdp6->inet6;
873 newdp = dccp_sk(newsk);
874 newnp = inet6_sk(newsk);
876 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
878 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
879 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
880 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
881 newsk->sk_bound_dev_if = ireq6->iif;
883 /* Now IPv6 options...
885 First: no IPv4 options.
887 newinet->opt = NULL;
889 /* Clone RX bits */
890 newnp->rxopt.all = np->rxopt.all;
892 /* Clone pktoptions received with SYN */
893 newnp->pktoptions = NULL;
894 if (ireq6->pktopts != NULL) {
895 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
896 kfree_skb(ireq6->pktopts);
897 ireq6->pktopts = NULL;
898 if (newnp->pktoptions)
899 skb_set_owner_r(newnp->pktoptions, newsk);
901 newnp->opt = NULL;
902 newnp->mcast_oif = inet6_iif(skb);
903 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
906 * Clone native IPv6 options from listening socket (if any)
908 * Yes, keeping reference count would be much more clever, but we make
909 * one more one thing there: reattach optmem to newsk.
911 if (opt != NULL) {
912 newnp->opt = ipv6_dup_options(newsk, opt);
913 if (opt != np->opt)
914 sock_kfree_s(sk, opt, opt->tot_len);
917 inet_csk(newsk)->icsk_ext_hdr_len = 0;
918 if (newnp->opt != NULL)
919 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
920 newnp->opt->opt_flen);
922 dccp_sync_mss(newsk, dst_mtu(dst));
924 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
926 __inet6_hash(&dccp_hashinfo, newsk);
927 inet_inherit_port(&dccp_hashinfo, sk, newsk);
929 return newsk;
931 out_overflow:
932 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
933 out:
934 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
935 if (opt != NULL && opt != np->opt)
936 sock_kfree_s(sk, opt, opt->tot_len);
937 dst_release(dst);
938 return NULL;
941 /* The socket must have it's spinlock held when we get
942 * here.
944 * We have a potential double-lock case here, so even when
945 * doing backlog processing we use the BH locking scheme.
946 * This is because we cannot sleep with the original spinlock
947 * held.
949 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
951 struct ipv6_pinfo *np = inet6_sk(sk);
952 struct sk_buff *opt_skb = NULL;
954 /* Imagine: socket is IPv6. IPv4 packet arrives,
955 goes to IPv4 receive handler and backlogged.
956 From backlog it always goes here. Kerboom...
957 Fortunately, dccp_rcv_established and rcv_established
958 handle them correctly, but it is not case with
959 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
962 if (skb->protocol == htons(ETH_P_IP))
963 return dccp_v4_do_rcv(sk, skb);
965 if (sk_filter(sk, skb, 0))
966 goto discard;
969 * socket locking is here for SMP purposes as backlog rcv is currently
970 * called with bh processing disabled.
973 /* Do Stevens' IPV6_PKTOPTIONS.
975 Yes, guys, it is the only place in our code, where we
976 may make it not affecting IPv4.
977 The rest of code is protocol independent,
978 and I do not like idea to uglify IPv4.
980 Actually, all the idea behind IPV6_PKTOPTIONS
981 looks not very well thought. For now we latch
982 options, received in the last packet, enqueued
983 by tcp. Feel free to propose better solution.
984 --ANK (980728)
986 if (np->rxopt.all)
987 opt_skb = skb_clone(skb, GFP_ATOMIC);
989 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
990 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
991 goto reset;
992 return 0;
995 if (sk->sk_state == DCCP_LISTEN) {
996 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
998 if (nsk == NULL)
999 goto discard;
1001 * Queue it on the new socket if the new socket is active,
1002 * otherwise we just shortcircuit this and continue with
1003 * the new socket..
1005 if (nsk != sk) {
1006 if (dccp_child_process(sk, nsk, skb))
1007 goto reset;
1008 if (opt_skb != NULL)
1009 __kfree_skb(opt_skb);
1010 return 0;
1014 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
1015 goto reset;
1016 return 0;
1018 reset:
1019 dccp_v6_ctl_send_reset(skb);
1020 discard:
1021 if (opt_skb != NULL)
1022 __kfree_skb(opt_skb);
1023 kfree_skb(skb);
1024 return 0;
1027 static int dccp_v6_rcv(struct sk_buff **pskb)
1029 const struct dccp_hdr *dh;
1030 struct sk_buff *skb = *pskb;
1031 struct sock *sk;
1033 /* Step 1: Check header basics: */
1035 if (dccp_invalid_packet(skb))
1036 goto discard_it;
1038 dh = dccp_hdr(skb);
1040 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
1041 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
1043 if (dccp_packet_without_ack(skb))
1044 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
1045 else
1046 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
1048 /* Step 2:
1049 * Look up flow ID in table and get corresponding socket */
1050 sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
1051 dh->dccph_sport,
1052 &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
1053 inet6_iif(skb));
1055 * Step 2:
1056 * If no socket ...
1057 * Generate Reset(No Connection) unless P.type == Reset
1058 * Drop packet and return
1060 if (sk == NULL)
1061 goto no_dccp_socket;
1064 * Step 2:
1065 * ... or S.state == TIMEWAIT,
1066 * Generate Reset(No Connection) unless P.type == Reset
1067 * Drop packet and return
1069 if (sk->sk_state == DCCP_TIME_WAIT)
1070 goto do_time_wait;
1072 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1073 goto discard_and_relse;
1075 return sk_receive_skb(sk, skb) ? -1 : 0;
1077 no_dccp_socket:
1078 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1079 goto discard_it;
1081 * Step 2:
1082 * Generate Reset(No Connection) unless P.type == Reset
1083 * Drop packet and return
1085 if (dh->dccph_type != DCCP_PKT_RESET) {
1086 DCCP_SKB_CB(skb)->dccpd_reset_code =
1087 DCCP_RESET_CODE_NO_CONNECTION;
1088 dccp_v6_ctl_send_reset(skb);
1090 discard_it:
1093 * Discard frame
1096 kfree_skb(skb);
1097 return 0;
1099 discard_and_relse:
1100 sock_put(sk);
1101 goto discard_it;
1103 do_time_wait:
1104 inet_twsk_put((struct inet_timewait_sock *)sk);
1105 goto no_dccp_socket;
1108 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1109 .queue_xmit = inet6_csk_xmit,
1110 .send_check = dccp_v6_send_check,
1111 .rebuild_header = inet6_sk_rebuild_header,
1112 .conn_request = dccp_v6_conn_request,
1113 .syn_recv_sock = dccp_v6_request_recv_sock,
1114 .net_header_len = sizeof(struct ipv6hdr),
1115 .setsockopt = ipv6_setsockopt,
1116 .getsockopt = ipv6_getsockopt,
1117 .addr2sockaddr = inet6_csk_addr2sockaddr,
1118 .sockaddr_len = sizeof(struct sockaddr_in6),
1119 #ifdef CONFIG_COMPAT
1120 .compat_setsockopt = compat_ipv6_setsockopt,
1121 .compat_getsockopt = compat_ipv6_getsockopt,
1122 #endif
1126 * DCCP over IPv4 via INET6 API
1128 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1129 .queue_xmit = ip_queue_xmit,
1130 .send_check = dccp_v4_send_check,
1131 .rebuild_header = inet_sk_rebuild_header,
1132 .conn_request = dccp_v6_conn_request,
1133 .syn_recv_sock = dccp_v6_request_recv_sock,
1134 .net_header_len = sizeof(struct iphdr),
1135 .setsockopt = ipv6_setsockopt,
1136 .getsockopt = ipv6_getsockopt,
1137 .addr2sockaddr = inet6_csk_addr2sockaddr,
1138 .sockaddr_len = sizeof(struct sockaddr_in6),
1139 #ifdef CONFIG_COMPAT
1140 .compat_setsockopt = compat_ipv6_setsockopt,
1141 .compat_getsockopt = compat_ipv6_getsockopt,
1142 #endif
1145 /* NOTE: A lot of things set to zero explicitly by call to
1146 * sk_alloc() so need not be done here.
1148 static int dccp_v6_init_sock(struct sock *sk)
1150 static __u8 dccp_v6_ctl_sock_initialized;
1151 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1153 if (err == 0) {
1154 if (unlikely(!dccp_v6_ctl_sock_initialized))
1155 dccp_v6_ctl_sock_initialized = 1;
1156 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1159 return err;
1162 static int dccp_v6_destroy_sock(struct sock *sk)
1164 dccp_destroy_sock(sk);
1165 return inet6_destroy_sock(sk);
1168 static struct proto dccp_v6_prot = {
1169 .name = "DCCPv6",
1170 .owner = THIS_MODULE,
1171 .close = dccp_close,
1172 .connect = dccp_v6_connect,
1173 .disconnect = dccp_disconnect,
1174 .ioctl = dccp_ioctl,
1175 .init = dccp_v6_init_sock,
1176 .setsockopt = dccp_setsockopt,
1177 .getsockopt = dccp_getsockopt,
1178 .sendmsg = dccp_sendmsg,
1179 .recvmsg = dccp_recvmsg,
1180 .backlog_rcv = dccp_v6_do_rcv,
1181 .hash = dccp_v6_hash,
1182 .unhash = dccp_unhash,
1183 .accept = inet_csk_accept,
1184 .get_port = dccp_v6_get_port,
1185 .shutdown = dccp_shutdown,
1186 .destroy = dccp_v6_destroy_sock,
1187 .orphan_count = &dccp_orphan_count,
1188 .max_header = MAX_DCCP_HEADER,
1189 .obj_size = sizeof(struct dccp6_sock),
1190 .rsk_prot = &dccp6_request_sock_ops,
1191 .twsk_prot = &dccp6_timewait_sock_ops,
1192 #ifdef CONFIG_COMPAT
1193 .compat_setsockopt = compat_dccp_setsockopt,
1194 .compat_getsockopt = compat_dccp_getsockopt,
1195 #endif
1198 static struct inet6_protocol dccp_v6_protocol = {
1199 .handler = dccp_v6_rcv,
1200 .err_handler = dccp_v6_err,
1201 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1204 static struct proto_ops inet6_dccp_ops = {
1205 .family = PF_INET6,
1206 .owner = THIS_MODULE,
1207 .release = inet6_release,
1208 .bind = inet6_bind,
1209 .connect = inet_stream_connect,
1210 .socketpair = sock_no_socketpair,
1211 .accept = inet_accept,
1212 .getname = inet6_getname,
1213 .poll = dccp_poll,
1214 .ioctl = inet6_ioctl,
1215 .listen = inet_dccp_listen,
1216 .shutdown = inet_shutdown,
1217 .setsockopt = sock_common_setsockopt,
1218 .getsockopt = sock_common_getsockopt,
1219 .sendmsg = inet_sendmsg,
1220 .recvmsg = sock_common_recvmsg,
1221 .mmap = sock_no_mmap,
1222 .sendpage = sock_no_sendpage,
1223 #ifdef CONFIG_COMPAT
1224 .compat_setsockopt = compat_sock_common_setsockopt,
1225 .compat_getsockopt = compat_sock_common_getsockopt,
1226 #endif
1229 static struct inet_protosw dccp_v6_protosw = {
1230 .type = SOCK_DCCP,
1231 .protocol = IPPROTO_DCCP,
1232 .prot = &dccp_v6_prot,
1233 .ops = &inet6_dccp_ops,
1234 .capability = -1,
1235 .flags = INET_PROTOSW_ICSK,
1238 static int __init dccp_v6_init(void)
1240 int err = proto_register(&dccp_v6_prot, 1);
1242 if (err != 0)
1243 goto out;
1245 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1246 if (err != 0)
1247 goto out_unregister_proto;
1249 inet6_register_protosw(&dccp_v6_protosw);
1251 err = inet_csk_ctl_sock_create(&dccp_v6_ctl_socket, PF_INET6,
1252 SOCK_DCCP, IPPROTO_DCCP);
1253 if (err != 0)
1254 goto out_unregister_protosw;
1255 out:
1256 return err;
1257 out_unregister_protosw:
1258 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1259 inet6_unregister_protosw(&dccp_v6_protosw);
1260 out_unregister_proto:
1261 proto_unregister(&dccp_v6_prot);
1262 goto out;
1265 static void __exit dccp_v6_exit(void)
1267 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1268 inet6_unregister_protosw(&dccp_v6_protosw);
1269 proto_unregister(&dccp_v6_prot);
1272 module_init(dccp_v6_init);
1273 module_exit(dccp_v6_exit);
1276 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1277 * values directly, Also cover the case where the protocol is not specified,
1278 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1280 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
1281 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
1282 MODULE_LICENSE("GPL");
1283 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1284 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");