Linux 2.6.19-rc6
[cris-mirror.git] / net / dccp / ipv6.c
blobeb0ff7ab05ed3a1d4062522cf6f58846c82e3fcd
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 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;
204 security_sk_classify_flow(sk, &fl);
206 if (np->opt != NULL && np->opt->srcrt != NULL) {
207 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
209 ipv6_addr_copy(&final, &fl.fl6_dst);
210 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
211 final_p = &final;
214 err = ip6_dst_lookup(sk, &dst, &fl);
215 if (err)
216 goto failure;
218 if (final_p)
219 ipv6_addr_copy(&fl.fl6_dst, final_p);
221 err = xfrm_lookup(&dst, &fl, sk, 0);
222 if (err < 0)
223 goto failure;
225 if (saddr == NULL) {
226 saddr = &fl.fl6_src;
227 ipv6_addr_copy(&np->rcv_saddr, saddr);
230 /* set the source address */
231 ipv6_addr_copy(&np->saddr, saddr);
232 inet->rcv_saddr = LOOPBACK4_IPV6;
234 __ip6_dst_store(sk, dst, NULL, NULL);
236 icsk->icsk_ext_hdr_len = 0;
237 if (np->opt != NULL)
238 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
239 np->opt->opt_nflen);
241 inet->dport = usin->sin6_port;
243 dccp_set_state(sk, DCCP_REQUESTING);
244 err = inet6_hash_connect(&dccp_death_row, sk);
245 if (err)
246 goto late_failure;
247 /* FIXME */
248 #if 0
249 dp->dccps_gar = secure_dccp_v6_sequence_number(np->saddr.s6_addr32,
250 np->daddr.s6_addr32,
251 inet->sport,
252 inet->dport);
253 #endif
254 err = dccp_connect(sk);
255 if (err)
256 goto late_failure;
258 return 0;
260 late_failure:
261 dccp_set_state(sk, DCCP_CLOSED);
262 __sk_dst_reset(sk);
263 failure:
264 inet->dport = 0;
265 sk->sk_route_caps = 0;
266 return err;
269 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
270 int type, int code, int offset, __be32 info)
272 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
273 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
274 struct ipv6_pinfo *np;
275 struct sock *sk;
276 int err;
277 __u64 seq;
279 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
280 &hdr->saddr, dh->dccph_sport, skb->dev->ifindex);
282 if (sk == NULL) {
283 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
284 return;
287 if (sk->sk_state == DCCP_TIME_WAIT) {
288 inet_twsk_put(inet_twsk(sk));
289 return;
292 bh_lock_sock(sk);
293 if (sock_owned_by_user(sk))
294 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
296 if (sk->sk_state == DCCP_CLOSED)
297 goto out;
299 np = inet6_sk(sk);
301 if (type == ICMPV6_PKT_TOOBIG) {
302 struct dst_entry *dst = NULL;
304 if (sock_owned_by_user(sk))
305 goto out;
306 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
307 goto out;
309 /* icmp should have updated the destination cache entry */
310 dst = __sk_dst_check(sk, np->dst_cookie);
311 if (dst == NULL) {
312 struct inet_sock *inet = inet_sk(sk);
313 struct flowi fl;
315 /* BUGGG_FUTURE: Again, it is not clear how
316 to handle rthdr case. Ignore this complexity
317 for now.
319 memset(&fl, 0, sizeof(fl));
320 fl.proto = IPPROTO_DCCP;
321 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
322 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
323 fl.oif = sk->sk_bound_dev_if;
324 fl.fl_ip_dport = inet->dport;
325 fl.fl_ip_sport = inet->sport;
326 security_sk_classify_flow(sk, &fl);
328 err = ip6_dst_lookup(sk, &dst, &fl);
329 if (err) {
330 sk->sk_err_soft = -err;
331 goto out;
334 err = xfrm_lookup(&dst, &fl, sk, 0);
335 if (err < 0) {
336 sk->sk_err_soft = -err;
337 goto out;
339 } else
340 dst_hold(dst);
342 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
343 dccp_sync_mss(sk, dst_mtu(dst));
344 } /* else let the usual retransmit timer handle it */
345 dst_release(dst);
346 goto out;
349 icmpv6_err_convert(type, code, &err);
351 seq = DCCP_SKB_CB(skb)->dccpd_seq;
352 /* Might be for an request_sock */
353 switch (sk->sk_state) {
354 struct request_sock *req, **prev;
355 case DCCP_LISTEN:
356 if (sock_owned_by_user(sk))
357 goto out;
359 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
360 &hdr->daddr, &hdr->saddr,
361 inet6_iif(skb));
362 if (req == NULL)
363 goto out;
366 * ICMPs are not backlogged, hence we cannot get an established
367 * socket here.
369 BUG_TRAP(req->sk == NULL);
371 if (seq != dccp_rsk(req)->dreq_iss) {
372 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
373 goto out;
376 inet_csk_reqsk_queue_drop(sk, req, prev);
377 goto out;
379 case DCCP_REQUESTING:
380 case DCCP_RESPOND: /* Cannot happen.
381 It can, it SYNs are crossed. --ANK */
382 if (!sock_owned_by_user(sk)) {
383 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
384 sk->sk_err = err;
386 * Wake people up to see the error
387 * (see connect in sock.c)
389 sk->sk_error_report(sk);
390 dccp_done(sk);
391 } else
392 sk->sk_err_soft = err;
393 goto out;
396 if (!sock_owned_by_user(sk) && np->recverr) {
397 sk->sk_err = err;
398 sk->sk_error_report(sk);
399 } else
400 sk->sk_err_soft = err;
402 out:
403 bh_unlock_sock(sk);
404 sock_put(sk);
408 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
409 struct dst_entry *dst)
411 struct inet6_request_sock *ireq6 = inet6_rsk(req);
412 struct ipv6_pinfo *np = inet6_sk(sk);
413 struct sk_buff *skb;
414 struct ipv6_txoptions *opt = NULL;
415 struct in6_addr *final_p = NULL, final;
416 struct flowi fl;
417 int err = -1;
419 memset(&fl, 0, sizeof(fl));
420 fl.proto = IPPROTO_DCCP;
421 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
422 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
423 fl.fl6_flowlabel = 0;
424 fl.oif = ireq6->iif;
425 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
426 fl.fl_ip_sport = inet_sk(sk)->sport;
427 security_req_classify_flow(req, &fl);
429 if (dst == NULL) {
430 opt = np->opt;
431 if (opt == NULL &&
432 np->rxopt.bits.osrcrt == 2 &&
433 ireq6->pktopts) {
434 struct sk_buff *pktopts = ireq6->pktopts;
435 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
437 if (rxopt->srcrt)
438 opt = ipv6_invert_rthdr(sk,
439 (struct ipv6_rt_hdr *)(pktopts->nh.raw +
440 rxopt->srcrt));
443 if (opt != NULL && opt->srcrt != NULL) {
444 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
446 ipv6_addr_copy(&final, &fl.fl6_dst);
447 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
448 final_p = &final;
451 err = ip6_dst_lookup(sk, &dst, &fl);
452 if (err)
453 goto done;
455 if (final_p)
456 ipv6_addr_copy(&fl.fl6_dst, final_p);
458 err = xfrm_lookup(&dst, &fl, sk, 0);
459 if (err < 0)
460 goto done;
463 skb = dccp_make_response(sk, dst, req);
464 if (skb != NULL) {
465 struct dccp_hdr *dh = dccp_hdr(skb);
467 dh->dccph_checksum = dccp_v6_check(dh, skb->len,
468 &ireq6->loc_addr,
469 &ireq6->rmt_addr,
470 csum_partial((char *)dh,
471 skb->len,
472 skb->csum));
473 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
474 err = ip6_xmit(sk, skb, &fl, opt, 0);
475 if (err == NET_XMIT_CN)
476 err = 0;
479 done:
480 if (opt != NULL && opt != np->opt)
481 sock_kfree_s(sk, opt, opt->tot_len);
482 dst_release(dst);
483 return err;
486 static void dccp_v6_reqsk_destructor(struct request_sock *req)
488 if (inet6_rsk(req)->pktopts != NULL)
489 kfree_skb(inet6_rsk(req)->pktopts);
492 static struct request_sock_ops dccp6_request_sock_ops = {
493 .family = AF_INET6,
494 .obj_size = sizeof(struct dccp6_request_sock),
495 .rtx_syn_ack = dccp_v6_send_response,
496 .send_ack = dccp_v6_reqsk_send_ack,
497 .destructor = dccp_v6_reqsk_destructor,
498 .send_reset = dccp_v6_ctl_send_reset,
501 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
502 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
505 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
507 struct ipv6_pinfo *np = inet6_sk(sk);
508 struct dccp_hdr *dh = dccp_hdr(skb);
510 dh->dccph_checksum = csum_ipv6_magic(&np->saddr, &np->daddr,
511 len, IPPROTO_DCCP,
512 csum_partial((char *)dh,
513 dh->dccph_doff << 2,
514 skb->csum));
517 static void dccp_v6_ctl_send_reset(struct sk_buff *rxskb)
519 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
520 const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
521 sizeof(struct dccp_hdr_ext) +
522 sizeof(struct dccp_hdr_reset);
523 struct sk_buff *skb;
524 struct flowi fl;
525 u64 seqno;
527 if (rxdh->dccph_type == DCCP_PKT_RESET)
528 return;
530 if (!ipv6_unicast_destination(rxskb))
531 return;
533 skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
534 GFP_ATOMIC);
535 if (skb == NULL)
536 return;
538 skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
540 skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
541 dh = dccp_hdr(skb);
542 memset(dh, 0, dccp_hdr_reset_len);
544 /* Swap the send and the receive. */
545 dh->dccph_type = DCCP_PKT_RESET;
546 dh->dccph_sport = rxdh->dccph_dport;
547 dh->dccph_dport = rxdh->dccph_sport;
548 dh->dccph_doff = dccp_hdr_reset_len / 4;
549 dh->dccph_x = 1;
550 dccp_hdr_reset(skb)->dccph_reset_code =
551 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
553 /* See "8.3.1. Abnormal Termination" in RFC 4340 */
554 seqno = 0;
555 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
556 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
558 dccp_hdr_set_seq(dh, seqno);
559 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
560 DCCP_SKB_CB(rxskb)->dccpd_seq);
562 memset(&fl, 0, sizeof(fl));
563 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
564 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
565 dh->dccph_checksum = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
566 sizeof(*dh), IPPROTO_DCCP,
567 skb->csum);
568 fl.proto = IPPROTO_DCCP;
569 fl.oif = inet6_iif(rxskb);
570 fl.fl_ip_dport = dh->dccph_dport;
571 fl.fl_ip_sport = dh->dccph_sport;
572 security_skb_classify_flow(rxskb, &fl);
574 /* sk = NULL, but it is safe for now. RST socket required. */
575 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
576 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
577 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
578 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
579 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
580 return;
584 kfree_skb(skb);
587 static void dccp_v6_reqsk_send_ack(struct sk_buff *rxskb,
588 struct request_sock *req)
590 struct flowi fl;
591 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
592 const u32 dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
593 sizeof(struct dccp_hdr_ext) +
594 sizeof(struct dccp_hdr_ack_bits);
595 struct sk_buff *skb;
597 skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
598 GFP_ATOMIC);
599 if (skb == NULL)
600 return;
602 skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
604 skb->h.raw = skb_push(skb, dccp_hdr_ack_len);
605 dh = dccp_hdr(skb);
606 memset(dh, 0, dccp_hdr_ack_len);
608 /* Build DCCP header and checksum it. */
609 dh->dccph_type = DCCP_PKT_ACK;
610 dh->dccph_sport = rxdh->dccph_dport;
611 dh->dccph_dport = rxdh->dccph_sport;
612 dh->dccph_doff = dccp_hdr_ack_len / 4;
613 dh->dccph_x = 1;
615 dccp_hdr_set_seq(dh, DCCP_SKB_CB(rxskb)->dccpd_ack_seq);
616 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
617 DCCP_SKB_CB(rxskb)->dccpd_seq);
619 memset(&fl, 0, sizeof(fl));
620 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
621 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
623 /* FIXME: calculate checksum, IPv4 also should... */
625 fl.proto = IPPROTO_DCCP;
626 fl.oif = inet6_iif(rxskb);
627 fl.fl_ip_dport = dh->dccph_dport;
628 fl.fl_ip_sport = dh->dccph_sport;
629 security_req_classify_flow(req, &fl);
631 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
632 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
633 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
634 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
635 return;
639 kfree_skb(skb);
642 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
644 const struct dccp_hdr *dh = dccp_hdr(skb);
645 const struct ipv6hdr *iph = skb->nh.ipv6h;
646 struct sock *nsk;
647 struct request_sock **prev;
648 /* Find possible connection requests. */
649 struct request_sock *req = inet6_csk_search_req(sk, &prev,
650 dh->dccph_sport,
651 &iph->saddr,
652 &iph->daddr,
653 inet6_iif(skb));
654 if (req != NULL)
655 return dccp_check_req(sk, skb, req, prev);
657 nsk = __inet6_lookup_established(&dccp_hashinfo,
658 &iph->saddr, dh->dccph_sport,
659 &iph->daddr, ntohs(dh->dccph_dport),
660 inet6_iif(skb));
661 if (nsk != NULL) {
662 if (nsk->sk_state != DCCP_TIME_WAIT) {
663 bh_lock_sock(nsk);
664 return nsk;
666 inet_twsk_put(inet_twsk(nsk));
667 return NULL;
670 return sk;
673 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
675 struct dccp_sock dp;
676 struct request_sock *req;
677 struct dccp_request_sock *dreq;
678 struct inet6_request_sock *ireq6;
679 struct ipv6_pinfo *np = inet6_sk(sk);
680 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
681 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
682 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
684 if (skb->protocol == htons(ETH_P_IP))
685 return dccp_v4_conn_request(sk, skb);
687 if (!ipv6_unicast_destination(skb))
688 goto drop;
690 if (dccp_bad_service_code(sk, service)) {
691 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
692 goto drop;
695 * There are no SYN attacks on IPv6, yet...
697 if (inet_csk_reqsk_queue_is_full(sk))
698 goto drop;
700 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
701 goto drop;
703 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
704 if (req == NULL)
705 goto drop;
707 /* FIXME: process options */
709 dccp_openreq_init(req, &dp, skb);
711 if (security_inet_conn_request(sk, skb, req))
712 goto drop_and_free;
714 ireq6 = inet6_rsk(req);
715 ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
716 ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
717 req->rcv_wnd = dccp_feat_default_sequence_window;
718 ireq6->pktopts = NULL;
720 if (ipv6_opt_accepted(sk, skb) ||
721 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
722 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
723 atomic_inc(&skb->users);
724 ireq6->pktopts = skb;
726 ireq6->iif = sk->sk_bound_dev_if;
728 /* So that link locals have meaning */
729 if (!sk->sk_bound_dev_if &&
730 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
731 ireq6->iif = inet6_iif(skb);
734 * Step 3: Process LISTEN state
736 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
738 * In fact we defer setting S.GSR, S.SWL, S.SWH to
739 * dccp_create_openreq_child.
741 dreq = dccp_rsk(req);
742 dreq->dreq_isr = dcb->dccpd_seq;
743 dreq->dreq_iss = dccp_v6_init_sequence(sk, skb);
744 dreq->dreq_service = service;
746 if (dccp_v6_send_response(sk, req, NULL))
747 goto drop_and_free;
749 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
750 return 0;
752 drop_and_free:
753 reqsk_free(req);
754 drop:
755 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
756 dcb->dccpd_reset_code = reset_code;
757 return -1;
760 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
761 struct sk_buff *skb,
762 struct request_sock *req,
763 struct dst_entry *dst)
765 struct inet6_request_sock *ireq6 = inet6_rsk(req);
766 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
767 struct inet_sock *newinet;
768 struct dccp_sock *newdp;
769 struct dccp6_sock *newdp6;
770 struct sock *newsk;
771 struct ipv6_txoptions *opt;
773 if (skb->protocol == htons(ETH_P_IP)) {
775 * v6 mapped
777 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
778 if (newsk == NULL)
779 return NULL;
781 newdp6 = (struct dccp6_sock *)newsk;
782 newdp = dccp_sk(newsk);
783 newinet = inet_sk(newsk);
784 newinet->pinet6 = &newdp6->inet6;
785 newnp = inet6_sk(newsk);
787 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
789 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
790 newinet->daddr);
792 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
793 newinet->saddr);
795 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
797 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
798 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
799 newnp->pktoptions = NULL;
800 newnp->opt = NULL;
801 newnp->mcast_oif = inet6_iif(skb);
802 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
805 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
806 * here, dccp_create_openreq_child now does this for us, see the comment in
807 * that function for the gory details. -acme
810 /* It is tricky place. Until this moment IPv4 tcp
811 worked with IPv6 icsk.icsk_af_ops.
812 Sync it now.
814 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
816 return newsk;
819 opt = np->opt;
821 if (sk_acceptq_is_full(sk))
822 goto out_overflow;
824 if (np->rxopt.bits.osrcrt == 2 && opt == NULL && ireq6->pktopts) {
825 const struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
827 if (rxopt->srcrt)
828 opt = ipv6_invert_rthdr(sk,
829 (struct ipv6_rt_hdr *)(ireq6->pktopts->nh.raw +
830 rxopt->srcrt));
833 if (dst == NULL) {
834 struct in6_addr *final_p = NULL, final;
835 struct flowi fl;
837 memset(&fl, 0, sizeof(fl));
838 fl.proto = IPPROTO_DCCP;
839 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
840 if (opt != NULL && opt->srcrt != NULL) {
841 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
843 ipv6_addr_copy(&final, &fl.fl6_dst);
844 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
845 final_p = &final;
847 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
848 fl.oif = sk->sk_bound_dev_if;
849 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
850 fl.fl_ip_sport = inet_sk(sk)->sport;
851 security_sk_classify_flow(sk, &fl);
853 if (ip6_dst_lookup(sk, &dst, &fl))
854 goto out;
856 if (final_p)
857 ipv6_addr_copy(&fl.fl6_dst, final_p);
859 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
860 goto out;
863 newsk = dccp_create_openreq_child(sk, req, skb);
864 if (newsk == NULL)
865 goto out;
868 * No need to charge this sock to the relevant IPv6 refcnt debug socks
869 * count here, dccp_create_openreq_child now does this for us, see the
870 * comment in that function for the gory details. -acme
873 __ip6_dst_store(newsk, dst, NULL, NULL);
874 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
875 NETIF_F_TSO);
876 newdp6 = (struct dccp6_sock *)newsk;
877 newinet = inet_sk(newsk);
878 newinet->pinet6 = &newdp6->inet6;
879 newdp = dccp_sk(newsk);
880 newnp = inet6_sk(newsk);
882 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
884 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
885 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
886 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
887 newsk->sk_bound_dev_if = ireq6->iif;
889 /* Now IPv6 options...
891 First: no IPv4 options.
893 newinet->opt = NULL;
895 /* Clone RX bits */
896 newnp->rxopt.all = np->rxopt.all;
898 /* Clone pktoptions received with SYN */
899 newnp->pktoptions = NULL;
900 if (ireq6->pktopts != NULL) {
901 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
902 kfree_skb(ireq6->pktopts);
903 ireq6->pktopts = NULL;
904 if (newnp->pktoptions)
905 skb_set_owner_r(newnp->pktoptions, newsk);
907 newnp->opt = NULL;
908 newnp->mcast_oif = inet6_iif(skb);
909 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
912 * Clone native IPv6 options from listening socket (if any)
914 * Yes, keeping reference count would be much more clever, but we make
915 * one more one thing there: reattach optmem to newsk.
917 if (opt != NULL) {
918 newnp->opt = ipv6_dup_options(newsk, opt);
919 if (opt != np->opt)
920 sock_kfree_s(sk, opt, opt->tot_len);
923 inet_csk(newsk)->icsk_ext_hdr_len = 0;
924 if (newnp->opt != NULL)
925 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
926 newnp->opt->opt_flen);
928 dccp_sync_mss(newsk, dst_mtu(dst));
930 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
932 __inet6_hash(&dccp_hashinfo, newsk);
933 inet_inherit_port(&dccp_hashinfo, sk, newsk);
935 return newsk;
937 out_overflow:
938 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
939 out:
940 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
941 if (opt != NULL && opt != np->opt)
942 sock_kfree_s(sk, opt, opt->tot_len);
943 dst_release(dst);
944 return NULL;
947 /* The socket must have it's spinlock held when we get
948 * here.
950 * We have a potential double-lock case here, so even when
951 * doing backlog processing we use the BH locking scheme.
952 * This is because we cannot sleep with the original spinlock
953 * held.
955 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
957 struct ipv6_pinfo *np = inet6_sk(sk);
958 struct sk_buff *opt_skb = NULL;
960 /* Imagine: socket is IPv6. IPv4 packet arrives,
961 goes to IPv4 receive handler and backlogged.
962 From backlog it always goes here. Kerboom...
963 Fortunately, dccp_rcv_established and rcv_established
964 handle them correctly, but it is not case with
965 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
968 if (skb->protocol == htons(ETH_P_IP))
969 return dccp_v4_do_rcv(sk, skb);
971 if (sk_filter(sk, skb))
972 goto discard;
975 * socket locking is here for SMP purposes as backlog rcv is currently
976 * called with bh processing disabled.
979 /* Do Stevens' IPV6_PKTOPTIONS.
981 Yes, guys, it is the only place in our code, where we
982 may make it not affecting IPv4.
983 The rest of code is protocol independent,
984 and I do not like idea to uglify IPv4.
986 Actually, all the idea behind IPV6_PKTOPTIONS
987 looks not very well thought. For now we latch
988 options, received in the last packet, enqueued
989 by tcp. Feel free to propose better solution.
990 --ANK (980728)
992 if (np->rxopt.all)
993 opt_skb = skb_clone(skb, GFP_ATOMIC);
995 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
996 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
997 goto reset;
998 if (opt_skb) {
999 /* This is where we would goto ipv6_pktoptions. */
1000 __kfree_skb(opt_skb);
1002 return 0;
1005 if (sk->sk_state == DCCP_LISTEN) {
1006 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
1008 if (nsk == NULL)
1009 goto discard;
1011 * Queue it on the new socket if the new socket is active,
1012 * otherwise we just shortcircuit this and continue with
1013 * the new socket..
1015 if (nsk != sk) {
1016 if (dccp_child_process(sk, nsk, skb))
1017 goto reset;
1018 if (opt_skb != NULL)
1019 __kfree_skb(opt_skb);
1020 return 0;
1024 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
1025 goto reset;
1026 if (opt_skb) {
1027 /* This is where we would goto ipv6_pktoptions. */
1028 __kfree_skb(opt_skb);
1030 return 0;
1032 reset:
1033 dccp_v6_ctl_send_reset(skb);
1034 discard:
1035 if (opt_skb != NULL)
1036 __kfree_skb(opt_skb);
1037 kfree_skb(skb);
1038 return 0;
1041 static int dccp_v6_rcv(struct sk_buff **pskb)
1043 const struct dccp_hdr *dh;
1044 struct sk_buff *skb = *pskb;
1045 struct sock *sk;
1047 /* Step 1: Check header basics: */
1049 if (dccp_invalid_packet(skb))
1050 goto discard_it;
1052 dh = dccp_hdr(skb);
1054 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
1055 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
1057 if (dccp_packet_without_ack(skb))
1058 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
1059 else
1060 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
1062 /* Step 2:
1063 * Look up flow ID in table and get corresponding socket */
1064 sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
1065 dh->dccph_sport,
1066 &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
1067 inet6_iif(skb));
1069 * Step 2:
1070 * If no socket ...
1071 * Generate Reset(No Connection) unless P.type == Reset
1072 * Drop packet and return
1074 if (sk == NULL)
1075 goto no_dccp_socket;
1078 * Step 2:
1079 * ... or S.state == TIMEWAIT,
1080 * Generate Reset(No Connection) unless P.type == Reset
1081 * Drop packet and return
1083 if (sk->sk_state == DCCP_TIME_WAIT)
1084 goto do_time_wait;
1086 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1087 goto discard_and_relse;
1089 return sk_receive_skb(sk, skb) ? -1 : 0;
1091 no_dccp_socket:
1092 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1093 goto discard_it;
1095 * Step 2:
1096 * Generate Reset(No Connection) unless P.type == Reset
1097 * Drop packet and return
1099 if (dh->dccph_type != DCCP_PKT_RESET) {
1100 DCCP_SKB_CB(skb)->dccpd_reset_code =
1101 DCCP_RESET_CODE_NO_CONNECTION;
1102 dccp_v6_ctl_send_reset(skb);
1104 discard_it:
1107 * Discard frame
1110 kfree_skb(skb);
1111 return 0;
1113 discard_and_relse:
1114 sock_put(sk);
1115 goto discard_it;
1117 do_time_wait:
1118 inet_twsk_put(inet_twsk(sk));
1119 goto no_dccp_socket;
1122 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1123 .queue_xmit = inet6_csk_xmit,
1124 .send_check = dccp_v6_send_check,
1125 .rebuild_header = inet6_sk_rebuild_header,
1126 .conn_request = dccp_v6_conn_request,
1127 .syn_recv_sock = dccp_v6_request_recv_sock,
1128 .net_header_len = sizeof(struct ipv6hdr),
1129 .setsockopt = ipv6_setsockopt,
1130 .getsockopt = ipv6_getsockopt,
1131 .addr2sockaddr = inet6_csk_addr2sockaddr,
1132 .sockaddr_len = sizeof(struct sockaddr_in6),
1133 #ifdef CONFIG_COMPAT
1134 .compat_setsockopt = compat_ipv6_setsockopt,
1135 .compat_getsockopt = compat_ipv6_getsockopt,
1136 #endif
1140 * DCCP over IPv4 via INET6 API
1142 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1143 .queue_xmit = ip_queue_xmit,
1144 .send_check = dccp_v4_send_check,
1145 .rebuild_header = inet_sk_rebuild_header,
1146 .conn_request = dccp_v6_conn_request,
1147 .syn_recv_sock = dccp_v6_request_recv_sock,
1148 .net_header_len = sizeof(struct iphdr),
1149 .setsockopt = ipv6_setsockopt,
1150 .getsockopt = ipv6_getsockopt,
1151 .addr2sockaddr = inet6_csk_addr2sockaddr,
1152 .sockaddr_len = sizeof(struct sockaddr_in6),
1153 #ifdef CONFIG_COMPAT
1154 .compat_setsockopt = compat_ipv6_setsockopt,
1155 .compat_getsockopt = compat_ipv6_getsockopt,
1156 #endif
1159 /* NOTE: A lot of things set to zero explicitly by call to
1160 * sk_alloc() so need not be done here.
1162 static int dccp_v6_init_sock(struct sock *sk)
1164 static __u8 dccp_v6_ctl_sock_initialized;
1165 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1167 if (err == 0) {
1168 if (unlikely(!dccp_v6_ctl_sock_initialized))
1169 dccp_v6_ctl_sock_initialized = 1;
1170 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1173 return err;
1176 static int dccp_v6_destroy_sock(struct sock *sk)
1178 dccp_destroy_sock(sk);
1179 return inet6_destroy_sock(sk);
1182 static struct proto dccp_v6_prot = {
1183 .name = "DCCPv6",
1184 .owner = THIS_MODULE,
1185 .close = dccp_close,
1186 .connect = dccp_v6_connect,
1187 .disconnect = dccp_disconnect,
1188 .ioctl = dccp_ioctl,
1189 .init = dccp_v6_init_sock,
1190 .setsockopt = dccp_setsockopt,
1191 .getsockopt = dccp_getsockopt,
1192 .sendmsg = dccp_sendmsg,
1193 .recvmsg = dccp_recvmsg,
1194 .backlog_rcv = dccp_v6_do_rcv,
1195 .hash = dccp_v6_hash,
1196 .unhash = dccp_unhash,
1197 .accept = inet_csk_accept,
1198 .get_port = dccp_v6_get_port,
1199 .shutdown = dccp_shutdown,
1200 .destroy = dccp_v6_destroy_sock,
1201 .orphan_count = &dccp_orphan_count,
1202 .max_header = MAX_DCCP_HEADER,
1203 .obj_size = sizeof(struct dccp6_sock),
1204 .rsk_prot = &dccp6_request_sock_ops,
1205 .twsk_prot = &dccp6_timewait_sock_ops,
1206 #ifdef CONFIG_COMPAT
1207 .compat_setsockopt = compat_dccp_setsockopt,
1208 .compat_getsockopt = compat_dccp_getsockopt,
1209 #endif
1212 static struct inet6_protocol dccp_v6_protocol = {
1213 .handler = dccp_v6_rcv,
1214 .err_handler = dccp_v6_err,
1215 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1218 static struct proto_ops inet6_dccp_ops = {
1219 .family = PF_INET6,
1220 .owner = THIS_MODULE,
1221 .release = inet6_release,
1222 .bind = inet6_bind,
1223 .connect = inet_stream_connect,
1224 .socketpair = sock_no_socketpair,
1225 .accept = inet_accept,
1226 .getname = inet6_getname,
1227 .poll = dccp_poll,
1228 .ioctl = inet6_ioctl,
1229 .listen = inet_dccp_listen,
1230 .shutdown = inet_shutdown,
1231 .setsockopt = sock_common_setsockopt,
1232 .getsockopt = sock_common_getsockopt,
1233 .sendmsg = inet_sendmsg,
1234 .recvmsg = sock_common_recvmsg,
1235 .mmap = sock_no_mmap,
1236 .sendpage = sock_no_sendpage,
1237 #ifdef CONFIG_COMPAT
1238 .compat_setsockopt = compat_sock_common_setsockopt,
1239 .compat_getsockopt = compat_sock_common_getsockopt,
1240 #endif
1243 static struct inet_protosw dccp_v6_protosw = {
1244 .type = SOCK_DCCP,
1245 .protocol = IPPROTO_DCCP,
1246 .prot = &dccp_v6_prot,
1247 .ops = &inet6_dccp_ops,
1248 .capability = -1,
1249 .flags = INET_PROTOSW_ICSK,
1252 static int __init dccp_v6_init(void)
1254 int err = proto_register(&dccp_v6_prot, 1);
1256 if (err != 0)
1257 goto out;
1259 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1260 if (err != 0)
1261 goto out_unregister_proto;
1263 inet6_register_protosw(&dccp_v6_protosw);
1265 err = inet_csk_ctl_sock_create(&dccp_v6_ctl_socket, PF_INET6,
1266 SOCK_DCCP, IPPROTO_DCCP);
1267 if (err != 0)
1268 goto out_unregister_protosw;
1269 out:
1270 return err;
1271 out_unregister_protosw:
1272 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1273 inet6_unregister_protosw(&dccp_v6_protosw);
1274 out_unregister_proto:
1275 proto_unregister(&dccp_v6_prot);
1276 goto out;
1279 static void __exit dccp_v6_exit(void)
1281 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1282 inet6_unregister_protosw(&dccp_v6_protosw);
1283 proto_unregister(&dccp_v6_prot);
1286 module_init(dccp_v6_init);
1287 module_exit(dccp_v6_exit);
1290 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1291 * values directly, Also cover the case where the protocol is not specified,
1292 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1294 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
1295 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
1296 MODULE_LICENSE("GPL");
1297 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1298 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");