Merge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-bugfix
[linux-2.6/kvm.git] / net / dccp / ipv6.c
blob80c4d048869e10e2880b5dff67ea22ccd1d0e8f6
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 static void dccp_v6_ctl_send_reset(struct sk_buff *skb);
37 static void dccp_v6_reqsk_send_ack(struct sk_buff *skb,
38 struct request_sock *req);
39 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb);
41 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
43 static struct inet_connection_sock_af_ops dccp_ipv6_mapped;
44 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
46 static int dccp_v6_get_port(struct sock *sk, unsigned short snum)
48 return inet_csk_get_port(&dccp_hashinfo, sk, snum,
49 inet6_csk_bind_conflict);
52 static void dccp_v6_hash(struct sock *sk)
54 if (sk->sk_state != DCCP_CLOSED) {
55 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
56 dccp_prot.hash(sk);
57 return;
59 local_bh_disable();
60 __inet6_hash(&dccp_hashinfo, sk);
61 local_bh_enable();
65 static inline u16 dccp_v6_check(struct dccp_hdr *dh, int len,
66 struct in6_addr *saddr,
67 struct in6_addr *daddr,
68 unsigned long base)
70 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_DCCP, base);
73 static __u32 dccp_v6_init_sequence(struct sock *sk, struct sk_buff *skb)
75 const struct dccp_hdr *dh = dccp_hdr(skb);
77 if (skb->protocol == htons(ETH_P_IPV6))
78 return secure_tcpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
79 skb->nh.ipv6h->saddr.s6_addr32,
80 dh->dccph_dport,
81 dh->dccph_sport);
82 else
83 return secure_dccp_sequence_number(skb->nh.iph->daddr,
84 skb->nh.iph->saddr,
85 dh->dccph_dport,
86 dh->dccph_sport);
89 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
90 int addr_len)
92 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
93 struct inet_connection_sock *icsk = inet_csk(sk);
94 struct inet_sock *inet = inet_sk(sk);
95 struct ipv6_pinfo *np = inet6_sk(sk);
96 struct dccp_sock *dp = dccp_sk(sk);
97 struct in6_addr *saddr = NULL, *final_p = NULL, final;
98 struct flowi fl;
99 struct dst_entry *dst;
100 int addr_type;
101 int err;
103 dp->dccps_role = DCCP_ROLE_CLIENT;
105 if (addr_len < SIN6_LEN_RFC2133)
106 return -EINVAL;
108 if (usin->sin6_family != AF_INET6)
109 return -EAFNOSUPPORT;
111 memset(&fl, 0, sizeof(fl));
113 if (np->sndflow) {
114 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
115 IP6_ECN_flow_init(fl.fl6_flowlabel);
116 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
117 struct ip6_flowlabel *flowlabel;
118 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
119 if (flowlabel == NULL)
120 return -EINVAL;
121 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
122 fl6_sock_release(flowlabel);
127 * connect() to INADDR_ANY means loopback (BSD'ism).
130 if (ipv6_addr_any(&usin->sin6_addr))
131 usin->sin6_addr.s6_addr[15] = 0x1;
133 addr_type = ipv6_addr_type(&usin->sin6_addr);
135 if(addr_type & IPV6_ADDR_MULTICAST)
136 return -ENETUNREACH;
138 if (addr_type & IPV6_ADDR_LINKLOCAL) {
139 if (addr_len >= sizeof(struct sockaddr_in6) &&
140 usin->sin6_scope_id) {
141 /* If interface is set while binding, indices
142 * must coincide.
144 if (sk->sk_bound_dev_if &&
145 sk->sk_bound_dev_if != usin->sin6_scope_id)
146 return -EINVAL;
148 sk->sk_bound_dev_if = usin->sin6_scope_id;
151 /* Connect to link-local address requires an interface */
152 if (!sk->sk_bound_dev_if)
153 return -EINVAL;
156 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
157 np->flow_label = fl.fl6_flowlabel;
160 * 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));
181 if (err) {
182 icsk->icsk_ext_hdr_len = exthdrlen;
183 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
184 sk->sk_backlog_rcv = dccp_v6_do_rcv;
185 goto failure;
186 } else {
187 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
188 inet->saddr);
189 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
190 inet->rcv_saddr);
193 return err;
196 if (!ipv6_addr_any(&np->rcv_saddr))
197 saddr = &np->rcv_saddr;
199 fl.proto = IPPROTO_DCCP;
200 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
201 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
202 fl.oif = sk->sk_bound_dev_if;
203 fl.fl_ip_dport = usin->sin6_port;
204 fl.fl_ip_sport = inet->sport;
206 if (np->opt && np->opt->srcrt) {
207 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;
216 if (final_p)
217 ipv6_addr_copy(&fl.fl6_dst, final_p);
219 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
220 goto failure;
222 if (saddr == NULL) {
223 saddr = &fl.fl6_src;
224 ipv6_addr_copy(&np->rcv_saddr, saddr);
227 /* set the source address */
228 ipv6_addr_copy(&np->saddr, saddr);
229 inet->rcv_saddr = LOOPBACK4_IPV6;
231 ip6_dst_store(sk, dst, NULL);
233 icsk->icsk_ext_hdr_len = 0;
234 if (np->opt)
235 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
236 np->opt->opt_nflen);
238 inet->dport = usin->sin6_port;
240 dccp_set_state(sk, DCCP_REQUESTING);
241 err = inet6_hash_connect(&dccp_death_row, sk);
242 if (err)
243 goto late_failure;
244 /* FIXME */
245 #if 0
246 dp->dccps_gar = secure_dccp_v6_sequence_number(np->saddr.s6_addr32,
247 np->daddr.s6_addr32,
248 inet->sport,
249 inet->dport);
250 #endif
251 err = dccp_connect(sk);
252 if (err)
253 goto late_failure;
255 return 0;
257 late_failure:
258 dccp_set_state(sk, DCCP_CLOSED);
259 __sk_dst_reset(sk);
260 failure:
261 inet->dport = 0;
262 sk->sk_route_caps = 0;
263 return err;
266 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
267 int type, int code, int offset, __u32 info)
269 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
270 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
271 struct ipv6_pinfo *np;
272 struct sock *sk;
273 int err;
274 __u64 seq;
276 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
277 &hdr->saddr, dh->dccph_sport, skb->dev->ifindex);
279 if (sk == NULL) {
280 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
281 return;
284 if (sk->sk_state == DCCP_TIME_WAIT) {
285 inet_twsk_put((struct inet_timewait_sock *)sk);
286 return;
289 bh_lock_sock(sk);
290 if (sock_owned_by_user(sk))
291 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
293 if (sk->sk_state == DCCP_CLOSED)
294 goto out;
296 np = inet6_sk(sk);
298 if (type == ICMPV6_PKT_TOOBIG) {
299 struct dst_entry *dst = NULL;
301 if (sock_owned_by_user(sk))
302 goto out;
303 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
304 goto out;
306 /* icmp should have updated the destination cache entry */
307 dst = __sk_dst_check(sk, np->dst_cookie);
309 if (dst == NULL) {
310 struct inet_sock *inet = inet_sk(sk);
311 struct flowi fl;
313 /* BUGGG_FUTURE: Again, it is not clear how
314 to handle rthdr case. Ignore this complexity
315 for now.
317 memset(&fl, 0, sizeof(fl));
318 fl.proto = IPPROTO_DCCP;
319 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
320 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
321 fl.oif = sk->sk_bound_dev_if;
322 fl.fl_ip_dport = inet->dport;
323 fl.fl_ip_sport = inet->sport;
325 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
326 sk->sk_err_soft = -err;
327 goto out;
330 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
331 sk->sk_err_soft = -err;
332 goto out;
335 } else
336 dst_hold(dst);
338 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
339 dccp_sync_mss(sk, dst_mtu(dst));
340 } /* else let the usual retransmit timer handle it */
341 dst_release(dst);
342 goto out;
345 icmpv6_err_convert(type, code, &err);
347 seq = DCCP_SKB_CB(skb)->dccpd_seq;
348 /* Might be for an request_sock */
349 switch (sk->sk_state) {
350 struct request_sock *req, **prev;
351 case DCCP_LISTEN:
352 if (sock_owned_by_user(sk))
353 goto out;
355 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
356 &hdr->daddr, &hdr->saddr,
357 inet6_iif(skb));
358 if (!req)
359 goto out;
361 /* ICMPs are not backlogged, hence we cannot get
362 * an established socket here.
364 BUG_TRAP(req->sk == NULL);
366 if (seq != dccp_rsk(req)->dreq_iss) {
367 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
368 goto out;
371 inet_csk_reqsk_queue_drop(sk, req, prev);
372 goto out;
374 case DCCP_REQUESTING:
375 case DCCP_RESPOND: /* Cannot happen.
376 It can, it SYNs are crossed. --ANK */
377 if (!sock_owned_by_user(sk)) {
378 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
379 sk->sk_err = err;
381 * Wake people up to see the error
382 * (see connect in sock.c)
384 sk->sk_error_report(sk);
386 dccp_done(sk);
387 } else
388 sk->sk_err_soft = err;
389 goto out;
392 if (!sock_owned_by_user(sk) && np->recverr) {
393 sk->sk_err = err;
394 sk->sk_error_report(sk);
395 } else
396 sk->sk_err_soft = err;
398 out:
399 bh_unlock_sock(sk);
400 sock_put(sk);
404 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
405 struct dst_entry *dst)
407 struct inet6_request_sock *ireq6 = inet6_rsk(req);
408 struct ipv6_pinfo *np = inet6_sk(sk);
409 struct sk_buff *skb;
410 struct ipv6_txoptions *opt = NULL;
411 struct in6_addr *final_p = NULL, final;
412 struct flowi fl;
413 int err = -1;
415 memset(&fl, 0, sizeof(fl));
416 fl.proto = IPPROTO_DCCP;
417 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
418 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
419 fl.fl6_flowlabel = 0;
420 fl.oif = ireq6->iif;
421 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
422 fl.fl_ip_sport = inet_sk(sk)->sport;
424 if (dst == NULL) {
425 opt = np->opt;
426 if (opt == NULL &&
427 np->rxopt.bits.osrcrt == 2 &&
428 ireq6->pktopts) {
429 struct sk_buff *pktopts = ireq6->pktopts;
430 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
431 if (rxopt->srcrt)
432 opt = ipv6_invert_rthdr(sk,
433 (struct ipv6_rt_hdr *)(pktopts->nh.raw +
434 rxopt->srcrt));
437 if (opt && opt->srcrt) {
438 struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
439 ipv6_addr_copy(&final, &fl.fl6_dst);
440 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
441 final_p = &final;
444 err = ip6_dst_lookup(sk, &dst, &fl);
445 if (err)
446 goto done;
447 if (final_p)
448 ipv6_addr_copy(&fl.fl6_dst, final_p);
449 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
450 goto done;
453 skb = dccp_make_response(sk, dst, req);
454 if (skb != NULL) {
455 struct dccp_hdr *dh = dccp_hdr(skb);
456 dh->dccph_checksum = dccp_v6_check(dh, skb->len,
457 &ireq6->loc_addr,
458 &ireq6->rmt_addr,
459 csum_partial((char *)dh,
460 skb->len,
461 skb->csum));
462 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
463 err = ip6_xmit(sk, skb, &fl, opt, 0);
464 if (err == NET_XMIT_CN)
465 err = 0;
468 done:
469 if (opt && opt != np->opt)
470 sock_kfree_s(sk, opt, opt->tot_len);
471 dst_release(dst);
472 return err;
475 static void dccp_v6_reqsk_destructor(struct request_sock *req)
477 if (inet6_rsk(req)->pktopts != NULL)
478 kfree_skb(inet6_rsk(req)->pktopts);
481 static struct request_sock_ops dccp6_request_sock_ops = {
482 .family = AF_INET6,
483 .obj_size = sizeof(struct dccp6_request_sock),
484 .rtx_syn_ack = dccp_v6_send_response,
485 .send_ack = dccp_v6_reqsk_send_ack,
486 .destructor = dccp_v6_reqsk_destructor,
487 .send_reset = dccp_v6_ctl_send_reset,
490 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
491 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
494 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
496 struct ipv6_pinfo *np = inet6_sk(sk);
497 struct dccp_hdr *dh = dccp_hdr(skb);
499 dh->dccph_checksum = csum_ipv6_magic(&np->saddr, &np->daddr,
500 len, IPPROTO_DCCP,
501 csum_partial((char *)dh,
502 dh->dccph_doff << 2,
503 skb->csum));
506 static void dccp_v6_ctl_send_reset(struct sk_buff *rxskb)
508 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
509 const int dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
510 sizeof(struct dccp_hdr_ext) +
511 sizeof(struct dccp_hdr_reset);
512 struct sk_buff *skb;
513 struct flowi fl;
514 u64 seqno;
516 if (rxdh->dccph_type == DCCP_PKT_RESET)
517 return;
519 if (!ipv6_unicast_destination(rxskb))
520 return;
523 * We need to grab some memory, and put together an RST,
524 * and then put it into the queue to be sent.
527 skb = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) +
528 dccp_hdr_reset_len, GFP_ATOMIC);
529 if (skb == NULL)
530 return;
532 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr) +
533 dccp_hdr_reset_len);
535 skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
536 dh = dccp_hdr(skb);
537 memset(dh, 0, dccp_hdr_reset_len);
539 /* Swap the send and the receive. */
540 dh->dccph_type = DCCP_PKT_RESET;
541 dh->dccph_sport = rxdh->dccph_dport;
542 dh->dccph_dport = rxdh->dccph_sport;
543 dh->dccph_doff = dccp_hdr_reset_len / 4;
544 dh->dccph_x = 1;
545 dccp_hdr_reset(skb)->dccph_reset_code =
546 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
548 /* See "8.3.1. Abnormal Termination" in draft-ietf-dccp-spec-11 */
549 seqno = 0;
550 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
551 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
553 dccp_hdr_set_seq(dh, seqno);
554 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
555 DCCP_SKB_CB(rxskb)->dccpd_seq);
557 memset(&fl, 0, sizeof(fl));
558 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
559 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
560 dh->dccph_checksum = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
561 sizeof(*dh), IPPROTO_DCCP,
562 skb->csum);
563 fl.proto = IPPROTO_DCCP;
564 fl.oif = inet6_iif(rxskb);
565 fl.fl_ip_dport = dh->dccph_dport;
566 fl.fl_ip_sport = dh->dccph_sport;
568 /* sk = NULL, but it is safe for now. RST socket required. */
569 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
570 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
571 ip6_xmit(NULL, skb, &fl, NULL, 0);
572 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
573 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
574 return;
578 kfree_skb(skb);
581 static void dccp_v6_ctl_send_ack(struct sk_buff *rxskb)
583 struct flowi fl;
584 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
585 const int dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
586 sizeof(struct dccp_hdr_ext) +
587 sizeof(struct dccp_hdr_ack_bits);
588 struct sk_buff *skb;
590 skb = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) +
591 dccp_hdr_ack_len, GFP_ATOMIC);
592 if (skb == NULL)
593 return;
595 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr) +
596 dccp_hdr_ack_len);
598 skb->h.raw = skb_push(skb, dccp_hdr_ack_len);
599 dh = dccp_hdr(skb);
600 memset(dh, 0, dccp_hdr_ack_len);
602 /* Build DCCP header and checksum it. */
603 dh->dccph_type = DCCP_PKT_ACK;
604 dh->dccph_sport = rxdh->dccph_dport;
605 dh->dccph_dport = rxdh->dccph_sport;
606 dh->dccph_doff = dccp_hdr_ack_len / 4;
607 dh->dccph_x = 1;
609 dccp_hdr_set_seq(dh, DCCP_SKB_CB(rxskb)->dccpd_ack_seq);
610 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
611 DCCP_SKB_CB(rxskb)->dccpd_seq);
613 memset(&fl, 0, sizeof(fl));
614 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
615 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
617 /* FIXME: calculate checksum, IPv4 also should... */
619 fl.proto = IPPROTO_DCCP;
620 fl.oif = inet6_iif(rxskb);
621 fl.fl_ip_dport = dh->dccph_dport;
622 fl.fl_ip_sport = dh->dccph_sport;
624 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
625 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
626 ip6_xmit(NULL, skb, &fl, NULL, 0);
627 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
628 return;
632 kfree_skb(skb);
635 static void dccp_v6_reqsk_send_ack(struct sk_buff *skb,
636 struct request_sock *req)
638 dccp_v6_ctl_send_ack(skb);
641 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
643 const struct dccp_hdr *dh = dccp_hdr(skb);
644 const struct ipv6hdr *iph = skb->nh.ipv6h;
645 struct sock *nsk;
646 struct request_sock **prev;
647 /* Find possible connection requests. */
648 struct request_sock *req = inet6_csk_search_req(sk, &prev,
649 dh->dccph_sport,
650 &iph->saddr,
651 &iph->daddr,
652 inet6_iif(skb));
653 if (req != NULL)
654 return dccp_check_req(sk, skb, req, prev);
656 nsk = __inet6_lookup_established(&dccp_hashinfo,
657 &iph->saddr, dh->dccph_sport,
658 &iph->daddr, ntohs(dh->dccph_dport),
659 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((struct inet_timewait_sock *)nsk);
667 return NULL;
670 return sk;
673 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
675 struct inet_request_sock *ireq;
676 struct dccp_sock dp;
677 struct request_sock *req;
678 struct dccp_request_sock *dreq;
679 struct inet6_request_sock *ireq6;
680 struct ipv6_pinfo *np = inet6_sk(sk);
681 const __u32 service = dccp_hdr_request(skb)->dccph_req_service;
682 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
683 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
685 if (skb->protocol == htons(ETH_P_IP))
686 return dccp_v4_conn_request(sk, skb);
688 if (!ipv6_unicast_destination(skb))
689 goto drop;
691 if (dccp_bad_service_code(sk, service)) {
692 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
693 goto drop;
696 * There are no SYN attacks on IPv6, yet...
698 if (inet_csk_reqsk_queue_is_full(sk))
699 goto drop;
701 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
702 goto drop;
704 req = inet6_reqsk_alloc(sk->sk_prot->rsk_prot);
705 if (req == NULL)
706 goto drop;
708 /* FIXME: process options */
710 dccp_openreq_init(req, &dp, skb);
712 ireq6 = inet6_rsk(req);
713 ireq = inet_rsk(req);
714 ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
715 ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
716 req->rcv_wnd = 100; /* Fake, option parsing will get the
717 right value */
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
778 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
779 if (newsk == NULL)
780 return NULL;
782 newdp6 = (struct dccp6_sock *)newsk;
783 newdp = dccp_sk(newsk);
784 newinet = inet_sk(newsk);
785 newinet->pinet6 = &newdp6->inet6;
786 newnp = inet6_sk(newsk);
788 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
790 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
791 newinet->daddr);
793 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
794 newinet->saddr);
796 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
798 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
799 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
800 newnp->pktoptions = NULL;
801 newnp->opt = NULL;
802 newnp->mcast_oif = inet6_iif(skb);
803 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
806 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
807 * here, dccp_create_openreq_child now does this for us, see the comment in
808 * that function for the gory details. -acme
811 /* It is tricky place. Until this moment IPv4 tcp
812 worked with IPv6 icsk.icsk_af_ops.
813 Sync it now.
815 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
817 return newsk;
820 opt = np->opt;
822 if (sk_acceptq_is_full(sk))
823 goto out_overflow;
825 if (np->rxopt.bits.osrcrt == 2 &&
826 opt == NULL && ireq6->pktopts) {
827 struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
828 if (rxopt->srcrt)
829 opt = ipv6_invert_rthdr(sk,
830 (struct ipv6_rt_hdr *)(ireq6->pktopts->nh.raw +
831 rxopt->srcrt));
834 if (dst == NULL) {
835 struct in6_addr *final_p = NULL, final;
836 struct flowi fl;
838 memset(&fl, 0, sizeof(fl));
839 fl.proto = IPPROTO_DCCP;
840 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
841 if (opt && opt->srcrt) {
842 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;
852 if (ip6_dst_lookup(sk, &dst, &fl))
853 goto out;
855 if (final_p)
856 ipv6_addr_copy(&fl.fl6_dst, final_p);
858 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
859 goto out;
862 newsk = dccp_create_openreq_child(sk, req, skb);
863 if (newsk == NULL)
864 goto out;
867 * No need to charge this sock to the relevant IPv6 refcnt debug socks
868 * count here, dccp_create_openreq_child now does this for us, see the
869 * comment in that function for the gory details. -acme
872 ip6_dst_store(newsk, dst, NULL);
873 newsk->sk_route_caps = dst->dev->features &
874 ~(NETIF_F_IP_CSUM | 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;
911 /* Clone native IPv6 options from listening socket (if any)
913 Yes, keeping reference count would be much more clever,
914 but we make one more one thing there: reattach optmem
915 to newsk.
917 if (opt) {
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)
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 && 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, 0))
972 goto discard;
975 * socket locking is here for SMP purposes as backlog rcv
976 * is currently 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 return 0;
1001 if (sk->sk_state == DCCP_LISTEN) {
1002 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
1003 if (!nsk)
1004 goto discard;
1007 * Queue it on the new socket if the new socket is active,
1008 * otherwise we just shortcircuit this and continue with
1009 * the new socket..
1011 if(nsk != sk) {
1012 if (dccp_child_process(sk, nsk, skb))
1013 goto reset;
1014 if (opt_skb)
1015 __kfree_skb(opt_skb);
1016 return 0;
1020 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
1021 goto reset;
1022 return 0;
1024 reset:
1025 dccp_v6_ctl_send_reset(skb);
1026 discard:
1027 if (opt_skb)
1028 __kfree_skb(opt_skb);
1029 kfree_skb(skb);
1030 return 0;
1033 static int dccp_v6_rcv(struct sk_buff **pskb)
1035 const struct dccp_hdr *dh;
1036 struct sk_buff *skb = *pskb;
1037 struct sock *sk;
1039 /* Step 1: Check header basics: */
1041 if (dccp_invalid_packet(skb))
1042 goto discard_it;
1044 dh = dccp_hdr(skb);
1046 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
1047 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
1049 if (dccp_packet_without_ack(skb))
1050 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
1051 else
1052 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
1054 /* Step 2:
1055 * Look up flow ID in table and get corresponding socket */
1056 sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
1057 dh->dccph_sport,
1058 &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
1059 inet6_iif(skb));
1061 * Step 2:
1062 * If no socket ...
1063 * Generate Reset(No Connection) unless P.type == Reset
1064 * Drop packet and return
1066 if (sk == NULL)
1067 goto no_dccp_socket;
1070 * Step 2:
1071 * ... or S.state == TIMEWAIT,
1072 * Generate Reset(No Connection) unless P.type == Reset
1073 * Drop packet and return
1076 if (sk->sk_state == DCCP_TIME_WAIT)
1077 goto do_time_wait;
1079 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1080 goto discard_and_relse;
1082 return sk_receive_skb(sk, skb) ? -1 : 0;
1084 no_dccp_socket:
1085 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1086 goto discard_it;
1088 * Step 2:
1089 * Generate Reset(No Connection) unless P.type == Reset
1090 * Drop packet and return
1092 if (dh->dccph_type != DCCP_PKT_RESET) {
1093 DCCP_SKB_CB(skb)->dccpd_reset_code =
1094 DCCP_RESET_CODE_NO_CONNECTION;
1095 dccp_v6_ctl_send_reset(skb);
1097 discard_it:
1100 * Discard frame
1103 kfree_skb(skb);
1104 return 0;
1106 discard_and_relse:
1107 sock_put(sk);
1108 goto discard_it;
1110 do_time_wait:
1111 inet_twsk_put((struct inet_timewait_sock *)sk);
1112 goto no_dccp_socket;
1115 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1116 .queue_xmit = inet6_csk_xmit,
1117 .send_check = dccp_v6_send_check,
1118 .rebuild_header = inet6_sk_rebuild_header,
1119 .conn_request = dccp_v6_conn_request,
1120 .syn_recv_sock = dccp_v6_request_recv_sock,
1121 .net_header_len = sizeof(struct ipv6hdr),
1122 .setsockopt = ipv6_setsockopt,
1123 .getsockopt = ipv6_getsockopt,
1124 .addr2sockaddr = inet6_csk_addr2sockaddr,
1125 .sockaddr_len = sizeof(struct sockaddr_in6)
1129 * DCCP over IPv4 via INET6 API
1131 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1132 .queue_xmit = ip_queue_xmit,
1133 .send_check = dccp_v4_send_check,
1134 .rebuild_header = inet_sk_rebuild_header,
1135 .conn_request = dccp_v6_conn_request,
1136 .syn_recv_sock = dccp_v6_request_recv_sock,
1137 .net_header_len = sizeof(struct iphdr),
1138 .setsockopt = ipv6_setsockopt,
1139 .getsockopt = ipv6_getsockopt,
1140 .addr2sockaddr = inet6_csk_addr2sockaddr,
1141 .sockaddr_len = sizeof(struct sockaddr_in6)
1144 /* NOTE: A lot of things set to zero explicitly by call to
1145 * sk_alloc() so need not be done here.
1147 static int dccp_v6_init_sock(struct sock *sk)
1149 int err = dccp_v4_init_sock(sk);
1151 if (err == 0)
1152 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1154 return err;
1157 static int dccp_v6_destroy_sock(struct sock *sk)
1159 dccp_v4_destroy_sock(sk);
1160 return inet6_destroy_sock(sk);
1163 static struct proto dccp_v6_prot = {
1164 .name = "DCCPv6",
1165 .owner = THIS_MODULE,
1166 .close = dccp_close,
1167 .connect = dccp_v6_connect,
1168 .disconnect = dccp_disconnect,
1169 .ioctl = dccp_ioctl,
1170 .init = dccp_v6_init_sock,
1171 .setsockopt = dccp_setsockopt,
1172 .getsockopt = dccp_getsockopt,
1173 .sendmsg = dccp_sendmsg,
1174 .recvmsg = dccp_recvmsg,
1175 .backlog_rcv = dccp_v6_do_rcv,
1176 .hash = dccp_v6_hash,
1177 .unhash = dccp_unhash,
1178 .accept = inet_csk_accept,
1179 .get_port = dccp_v6_get_port,
1180 .shutdown = dccp_shutdown,
1181 .destroy = dccp_v6_destroy_sock,
1182 .orphan_count = &dccp_orphan_count,
1183 .max_header = MAX_DCCP_HEADER,
1184 .obj_size = sizeof(struct dccp6_sock),
1185 .rsk_prot = &dccp6_request_sock_ops,
1186 .twsk_prot = &dccp6_timewait_sock_ops,
1189 static struct inet6_protocol dccp_v6_protocol = {
1190 .handler = dccp_v6_rcv,
1191 .err_handler = dccp_v6_err,
1192 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1195 static struct proto_ops inet6_dccp_ops = {
1196 .family = PF_INET6,
1197 .owner = THIS_MODULE,
1198 .release = inet6_release,
1199 .bind = inet6_bind,
1200 .connect = inet_stream_connect,
1201 .socketpair = sock_no_socketpair,
1202 .accept = inet_accept,
1203 .getname = inet6_getname,
1204 .poll = dccp_poll,
1205 .ioctl = inet6_ioctl,
1206 .listen = inet_dccp_listen,
1207 .shutdown = inet_shutdown,
1208 .setsockopt = sock_common_setsockopt,
1209 .getsockopt = sock_common_getsockopt,
1210 .sendmsg = inet_sendmsg,
1211 .recvmsg = sock_common_recvmsg,
1212 .mmap = sock_no_mmap,
1213 .sendpage = sock_no_sendpage,
1216 static struct inet_protosw dccp_v6_protosw = {
1217 .type = SOCK_DCCP,
1218 .protocol = IPPROTO_DCCP,
1219 .prot = &dccp_v6_prot,
1220 .ops = &inet6_dccp_ops,
1221 .capability = -1,
1222 .flags = INET_PROTOSW_ICSK,
1225 static int __init dccp_v6_init(void)
1227 int err = proto_register(&dccp_v6_prot, 1);
1229 if (err != 0)
1230 goto out;
1232 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1233 if (err != 0)
1234 goto out_unregister_proto;
1236 inet6_register_protosw(&dccp_v6_protosw);
1237 out:
1238 return err;
1239 out_unregister_proto:
1240 proto_unregister(&dccp_v6_prot);
1241 goto out;
1244 static void __exit dccp_v6_exit(void)
1246 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1247 inet6_unregister_protosw(&dccp_v6_protosw);
1248 proto_unregister(&dccp_v6_prot);
1251 module_init(dccp_v6_init);
1252 module_exit(dccp_v6_exit);
1255 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1256 * values directly, Also cover the case where the protocol is not specified,
1257 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1259 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
1260 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
1261 MODULE_LICENSE("GPL");
1262 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1263 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");