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/inet6_connection_sock.h>
24 #include <net/inet6_hashtables.h>
25 #include <net/ip6_route.h>
27 #include <net/protocol.h>
28 #include <net/transp_v6.h>
34 static void dccp_v6_ctl_send_reset(struct sk_buff *skb);
35 static void dccp_v6_reqsk_send_ack(struct sk_buff *skb,
36 struct request_sock *req);
37 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb);
39 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
41 static struct inet_connection_sock_af_ops dccp_ipv6_mapped;
42 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
44 static int dccp_v6_get_port(struct sock *sk, unsigned short snum)
46 return inet_csk_get_port(&dccp_hashinfo, sk, snum,
47 inet6_csk_bind_conflict);
50 static void dccp_v6_hash(struct sock *sk)
52 if (sk->sk_state != DCCP_CLOSED) {
53 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
58 __inet6_hash(&dccp_hashinfo, sk);
63 static inline u16 dccp_v6_check(struct dccp_hdr *dh, int len,
64 struct in6_addr *saddr,
65 struct in6_addr *daddr,
68 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_DCCP, base);
71 static __u32 dccp_v6_init_sequence(struct sock *sk, struct sk_buff *skb)
73 const struct dccp_hdr *dh = dccp_hdr(skb);
75 if (skb->protocol == htons(ETH_P_IPV6))
76 return secure_tcpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
77 skb->nh.ipv6h->saddr.s6_addr32,
81 return secure_dccp_sequence_number(skb->nh.iph->daddr,
87 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
90 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
91 struct inet_connection_sock *icsk = inet_csk(sk);
92 struct inet_sock *inet = inet_sk(sk);
93 struct ipv6_pinfo *np = inet6_sk(sk);
94 struct dccp_sock *dp = dccp_sk(sk);
95 struct in6_addr *saddr = NULL, *final_p = NULL, final;
97 struct dst_entry *dst;
101 dp->dccps_role = DCCP_ROLE_CLIENT;
103 if (addr_len < SIN6_LEN_RFC2133)
106 if (usin->sin6_family != AF_INET6)
107 return -EAFNOSUPPORT;
109 memset(&fl, 0, sizeof(fl));
112 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
113 IP6_ECN_flow_init(fl.fl6_flowlabel);
114 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
115 struct ip6_flowlabel *flowlabel;
116 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
117 if (flowlabel == NULL)
119 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
120 fl6_sock_release(flowlabel);
125 * connect() to INADDR_ANY means loopback (BSD'ism).
128 if (ipv6_addr_any(&usin->sin6_addr))
129 usin->sin6_addr.s6_addr[15] = 0x1;
131 addr_type = ipv6_addr_type(&usin->sin6_addr);
133 if(addr_type & IPV6_ADDR_MULTICAST)
136 if (addr_type & IPV6_ADDR_LINKLOCAL) {
137 if (addr_len >= sizeof(struct sockaddr_in6) &&
138 usin->sin6_scope_id) {
139 /* If interface is set while binding, indices
142 if (sk->sk_bound_dev_if &&
143 sk->sk_bound_dev_if != usin->sin6_scope_id)
146 sk->sk_bound_dev_if = usin->sin6_scope_id;
149 /* Connect to link-local address requires an interface */
150 if (!sk->sk_bound_dev_if)
154 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
155 np->flow_label = fl.fl6_flowlabel;
161 if (addr_type == IPV6_ADDR_MAPPED) {
162 u32 exthdrlen = icsk->icsk_ext_hdr_len;
163 struct sockaddr_in sin;
165 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
167 if (__ipv6_only_sock(sk))
170 sin.sin_family = AF_INET;
171 sin.sin_port = usin->sin6_port;
172 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
174 icsk->icsk_af_ops = &dccp_ipv6_mapped;
175 sk->sk_backlog_rcv = dccp_v4_do_rcv;
177 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
180 icsk->icsk_ext_hdr_len = exthdrlen;
181 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
182 sk->sk_backlog_rcv = dccp_v6_do_rcv;
185 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
187 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
194 if (!ipv6_addr_any(&np->rcv_saddr))
195 saddr = &np->rcv_saddr;
197 fl.proto = IPPROTO_DCCP;
198 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
199 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
200 fl.oif = sk->sk_bound_dev_if;
201 fl.fl_ip_dport = usin->sin6_port;
202 fl.fl_ip_sport = inet->sport;
204 if (np->opt && np->opt->srcrt) {
205 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
206 ipv6_addr_copy(&final, &fl.fl6_dst);
207 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
211 err = ip6_dst_lookup(sk, &dst, &fl);
215 ipv6_addr_copy(&fl.fl6_dst, final_p);
217 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
222 ipv6_addr_copy(&np->rcv_saddr, saddr);
225 /* set the source address */
226 ipv6_addr_copy(&np->saddr, saddr);
227 inet->rcv_saddr = LOOPBACK4_IPV6;
229 ip6_dst_store(sk, dst, NULL);
231 icsk->icsk_ext_hdr_len = 0;
233 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
236 inet->dport = usin->sin6_port;
238 dccp_set_state(sk, DCCP_REQUESTING);
239 err = inet6_hash_connect(&dccp_death_row, sk);
244 dp->dccps_gar = secure_dccp_v6_sequence_number(np->saddr.s6_addr32,
249 err = dccp_connect(sk);
256 dccp_set_state(sk, DCCP_CLOSED);
260 sk->sk_route_caps = 0;
264 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
265 int type, int code, int offset, __u32 info)
267 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
268 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
269 struct ipv6_pinfo *np;
274 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
275 &hdr->saddr, dh->dccph_sport, skb->dev->ifindex);
278 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
282 if (sk->sk_state == DCCP_TIME_WAIT) {
283 inet_twsk_put((struct inet_timewait_sock *)sk);
288 if (sock_owned_by_user(sk))
289 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
291 if (sk->sk_state == DCCP_CLOSED)
296 if (type == ICMPV6_PKT_TOOBIG) {
297 struct dst_entry *dst = NULL;
299 if (sock_owned_by_user(sk))
301 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
304 /* icmp should have updated the destination cache entry */
305 dst = __sk_dst_check(sk, np->dst_cookie);
308 struct inet_sock *inet = inet_sk(sk);
311 /* BUGGG_FUTURE: Again, it is not clear how
312 to handle rthdr case. Ignore this complexity
315 memset(&fl, 0, sizeof(fl));
316 fl.proto = IPPROTO_DCCP;
317 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
318 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
319 fl.oif = sk->sk_bound_dev_if;
320 fl.fl_ip_dport = inet->dport;
321 fl.fl_ip_sport = inet->sport;
323 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
324 sk->sk_err_soft = -err;
328 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
329 sk->sk_err_soft = -err;
336 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
337 dccp_sync_mss(sk, dst_mtu(dst));
338 } /* else let the usual retransmit timer handle it */
343 icmpv6_err_convert(type, code, &err);
345 seq = DCCP_SKB_CB(skb)->dccpd_seq;
346 /* Might be for an request_sock */
347 switch (sk->sk_state) {
348 struct request_sock *req, **prev;
350 if (sock_owned_by_user(sk))
353 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
354 &hdr->daddr, &hdr->saddr,
359 /* ICMPs are not backlogged, hence we cannot get
360 * an established socket here.
362 BUG_TRAP(req->sk == NULL);
364 if (seq != dccp_rsk(req)->dreq_iss) {
365 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
369 inet_csk_reqsk_queue_drop(sk, req, prev);
372 case DCCP_REQUESTING:
373 case DCCP_RESPOND: /* Cannot happen.
374 It can, it SYNs are crossed. --ANK */
375 if (!sock_owned_by_user(sk)) {
376 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
379 * Wake people up to see the error
380 * (see connect in sock.c)
382 sk->sk_error_report(sk);
386 sk->sk_err_soft = err;
390 if (!sock_owned_by_user(sk) && np->recverr) {
392 sk->sk_error_report(sk);
394 sk->sk_err_soft = err;
402 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
403 struct dst_entry *dst)
405 struct inet6_request_sock *ireq6 = inet6_rsk(req);
406 struct ipv6_pinfo *np = inet6_sk(sk);
408 struct ipv6_txoptions *opt = NULL;
409 struct in6_addr *final_p = NULL, final;
413 memset(&fl, 0, sizeof(fl));
414 fl.proto = IPPROTO_DCCP;
415 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
416 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
417 fl.fl6_flowlabel = 0;
419 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
420 fl.fl_ip_sport = inet_sk(sk)->sport;
425 np->rxopt.bits.osrcrt == 2 &&
427 struct sk_buff *pktopts = ireq6->pktopts;
428 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
430 opt = ipv6_invert_rthdr(sk,
431 (struct ipv6_rt_hdr *)(pktopts->nh.raw +
435 if (opt && opt->srcrt) {
436 struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
437 ipv6_addr_copy(&final, &fl.fl6_dst);
438 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
442 err = ip6_dst_lookup(sk, &dst, &fl);
446 ipv6_addr_copy(&fl.fl6_dst, final_p);
447 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
451 skb = dccp_make_response(sk, dst, req);
453 struct dccp_hdr *dh = dccp_hdr(skb);
454 dh->dccph_checksum = dccp_v6_check(dh, skb->len,
457 csum_partial((char *)dh,
460 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
461 err = ip6_xmit(sk, skb, &fl, opt, 0);
462 if (err == NET_XMIT_CN)
467 if (opt && opt != np->opt)
468 sock_kfree_s(sk, opt, opt->tot_len);
472 static void dccp_v6_reqsk_destructor(struct request_sock *req)
474 if (inet6_rsk(req)->pktopts != NULL)
475 kfree_skb(inet6_rsk(req)->pktopts);
478 static struct request_sock_ops dccp6_request_sock_ops = {
480 .obj_size = sizeof(struct dccp6_request_sock),
481 .rtx_syn_ack = dccp_v6_send_response,
482 .send_ack = dccp_v6_reqsk_send_ack,
483 .destructor = dccp_v6_reqsk_destructor,
484 .send_reset = dccp_v6_ctl_send_reset,
487 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
488 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
491 static void dccp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
493 struct ipv6_pinfo *np = inet6_sk(sk);
494 struct dccp_hdr *dh = dccp_hdr(skb);
496 dh->dccph_checksum = csum_ipv6_magic(&np->saddr, &np->daddr,
498 csum_partial((char *)dh,
503 static void dccp_v6_ctl_send_reset(struct sk_buff *rxskb)
505 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
506 const int dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
507 sizeof(struct dccp_hdr_ext) +
508 sizeof(struct dccp_hdr_reset);
513 if (rxdh->dccph_type == DCCP_PKT_RESET)
516 if (!ipv6_unicast_destination(rxskb))
520 * We need to grab some memory, and put together an RST,
521 * and then put it into the queue to be sent.
524 skb = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) +
525 dccp_hdr_reset_len, GFP_ATOMIC);
529 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr) +
532 skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
534 memset(dh, 0, dccp_hdr_reset_len);
536 /* Swap the send and the receive. */
537 dh->dccph_type = DCCP_PKT_RESET;
538 dh->dccph_sport = rxdh->dccph_dport;
539 dh->dccph_dport = rxdh->dccph_sport;
540 dh->dccph_doff = dccp_hdr_reset_len / 4;
542 dccp_hdr_reset(skb)->dccph_reset_code =
543 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
545 /* See "8.3.1. Abnormal Termination" in draft-ietf-dccp-spec-11 */
547 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
548 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
550 dccp_hdr_set_seq(dh, seqno);
551 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
552 DCCP_SKB_CB(rxskb)->dccpd_seq);
554 memset(&fl, 0, sizeof(fl));
555 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
556 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
557 dh->dccph_checksum = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
558 sizeof(*dh), IPPROTO_DCCP,
560 fl.proto = IPPROTO_DCCP;
561 fl.oif = inet6_iif(rxskb);
562 fl.fl_ip_dport = dh->dccph_dport;
563 fl.fl_ip_sport = dh->dccph_sport;
565 /* sk = NULL, but it is safe for now. RST socket required. */
566 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
567 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
568 ip6_xmit(NULL, skb, &fl, NULL, 0);
569 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
570 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
578 static void dccp_v6_ctl_send_ack(struct sk_buff *rxskb)
581 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
582 const int dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
583 sizeof(struct dccp_hdr_ext) +
584 sizeof(struct dccp_hdr_ack_bits);
587 skb = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) +
588 dccp_hdr_ack_len, GFP_ATOMIC);
592 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr) +
595 skb->h.raw = skb_push(skb, dccp_hdr_ack_len);
597 memset(dh, 0, dccp_hdr_ack_len);
599 /* Build DCCP header and checksum it. */
600 dh->dccph_type = DCCP_PKT_ACK;
601 dh->dccph_sport = rxdh->dccph_dport;
602 dh->dccph_dport = rxdh->dccph_sport;
603 dh->dccph_doff = dccp_hdr_ack_len / 4;
606 dccp_hdr_set_seq(dh, DCCP_SKB_CB(rxskb)->dccpd_ack_seq);
607 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
608 DCCP_SKB_CB(rxskb)->dccpd_seq);
610 memset(&fl, 0, sizeof(fl));
611 ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
612 ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
614 /* FIXME: calculate checksum, IPv4 also should... */
616 fl.proto = IPPROTO_DCCP;
617 fl.oif = inet6_iif(rxskb);
618 fl.fl_ip_dport = dh->dccph_dport;
619 fl.fl_ip_sport = dh->dccph_sport;
621 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
622 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
623 ip6_xmit(NULL, skb, &fl, NULL, 0);
624 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
632 static void dccp_v6_reqsk_send_ack(struct sk_buff *skb,
633 struct request_sock *req)
635 dccp_v6_ctl_send_ack(skb);
638 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
640 const struct dccp_hdr *dh = dccp_hdr(skb);
641 const struct ipv6hdr *iph = skb->nh.ipv6h;
643 struct request_sock **prev;
644 /* Find possible connection requests. */
645 struct request_sock *req = inet6_csk_search_req(sk, &prev,
651 return dccp_check_req(sk, skb, req, prev);
653 nsk = __inet6_lookup_established(&dccp_hashinfo,
654 &iph->saddr, dh->dccph_sport,
655 &iph->daddr, ntohs(dh->dccph_dport),
659 if (nsk->sk_state != DCCP_TIME_WAIT) {
663 inet_twsk_put((struct inet_timewait_sock *)nsk);
670 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
672 struct inet_request_sock *ireq;
674 struct request_sock *req;
675 struct dccp_request_sock *dreq;
676 struct inet6_request_sock *ireq6;
677 struct ipv6_pinfo *np = inet6_sk(sk);
678 const __u32 service = dccp_hdr_request(skb)->dccph_req_service;
679 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
680 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
682 if (skb->protocol == htons(ETH_P_IP))
683 return dccp_v4_conn_request(sk, skb);
685 if (!ipv6_unicast_destination(skb))
688 if (dccp_bad_service_code(sk, service)) {
689 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
693 * There are no SYN attacks on IPv6, yet...
695 if (inet_csk_reqsk_queue_is_full(sk))
698 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
701 req = inet6_reqsk_alloc(sk->sk_prot->rsk_prot);
705 /* FIXME: process options */
707 dccp_openreq_init(req, &dp, skb);
709 ireq6 = inet6_rsk(req);
710 ireq = inet_rsk(req);
711 ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
712 ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
713 req->rcv_wnd = 100; /* Fake, option parsing will get the
715 ireq6->pktopts = NULL;
717 if (ipv6_opt_accepted(sk, skb) ||
718 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
719 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
720 atomic_inc(&skb->users);
721 ireq6->pktopts = skb;
723 ireq6->iif = sk->sk_bound_dev_if;
725 /* So that link locals have meaning */
726 if (!sk->sk_bound_dev_if &&
727 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
728 ireq6->iif = inet6_iif(skb);
731 * Step 3: Process LISTEN state
733 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
735 * In fact we defer setting S.GSR, S.SWL, S.SWH to
736 * dccp_create_openreq_child.
738 dreq = dccp_rsk(req);
739 dreq->dreq_isr = dcb->dccpd_seq;
740 dreq->dreq_iss = dccp_v6_init_sequence(sk, skb);
741 dreq->dreq_service = service;
743 if (dccp_v6_send_response(sk, req, NULL))
746 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
752 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
753 dcb->dccpd_reset_code = reset_code;
757 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
759 struct request_sock *req,
760 struct dst_entry *dst)
762 struct inet6_request_sock *ireq6 = inet6_rsk(req);
763 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
764 struct inet_sock *newinet;
765 struct dccp_sock *newdp;
766 struct dccp6_sock *newdp6;
768 struct ipv6_txoptions *opt;
770 if (skb->protocol == htons(ETH_P_IP)) {
775 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
779 newdp6 = (struct dccp6_sock *)newsk;
780 newdp = dccp_sk(newsk);
781 newinet = inet_sk(newsk);
782 newinet->pinet6 = &newdp6->inet6;
783 newnp = inet6_sk(newsk);
785 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
787 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
790 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
793 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
795 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
796 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
797 newnp->pktoptions = NULL;
799 newnp->mcast_oif = inet6_iif(skb);
800 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
803 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
804 * here, dccp_create_openreq_child now does this for us, see the comment in
805 * that function for the gory details. -acme
808 /* It is tricky place. Until this moment IPv4 tcp
809 worked with IPv6 icsk.icsk_af_ops.
812 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
819 if (sk_acceptq_is_full(sk))
822 if (np->rxopt.bits.osrcrt == 2 &&
823 opt == NULL && ireq6->pktopts) {
824 struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
826 opt = ipv6_invert_rthdr(sk,
827 (struct ipv6_rt_hdr *)(ireq6->pktopts->nh.raw +
832 struct in6_addr *final_p = NULL, final;
835 memset(&fl, 0, sizeof(fl));
836 fl.proto = IPPROTO_DCCP;
837 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
838 if (opt && opt->srcrt) {
839 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
840 ipv6_addr_copy(&final, &fl.fl6_dst);
841 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
844 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
845 fl.oif = sk->sk_bound_dev_if;
846 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
847 fl.fl_ip_sport = inet_sk(sk)->sport;
849 if (ip6_dst_lookup(sk, &dst, &fl))
853 ipv6_addr_copy(&fl.fl6_dst, final_p);
855 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
859 newsk = dccp_create_openreq_child(sk, req, skb);
864 * No need to charge this sock to the relevant IPv6 refcnt debug socks
865 * count here, dccp_create_openreq_child now does this for us, see the
866 * comment in that function for the gory details. -acme
869 ip6_dst_store(newsk, dst, NULL);
870 newsk->sk_route_caps = dst->dev->features &
871 ~(NETIF_F_IP_CSUM | NETIF_F_TSO);
873 newdp6 = (struct dccp6_sock *)newsk;
874 newinet = inet_sk(newsk);
875 newinet->pinet6 = &newdp6->inet6;
876 newdp = dccp_sk(newsk);
877 newnp = inet6_sk(newsk);
879 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
881 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
882 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
883 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
884 newsk->sk_bound_dev_if = ireq6->iif;
886 /* Now IPv6 options...
888 First: no IPv4 options.
893 newnp->rxopt.all = np->rxopt.all;
895 /* Clone pktoptions received with SYN */
896 newnp->pktoptions = NULL;
897 if (ireq6->pktopts != NULL) {
898 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
899 kfree_skb(ireq6->pktopts);
900 ireq6->pktopts = NULL;
901 if (newnp->pktoptions)
902 skb_set_owner_r(newnp->pktoptions, newsk);
905 newnp->mcast_oif = inet6_iif(skb);
906 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
908 /* Clone native IPv6 options from listening socket (if any)
910 Yes, keeping reference count would be much more clever,
911 but we make one more one thing there: reattach optmem
915 newnp->opt = ipv6_dup_options(newsk, opt);
917 sock_kfree_s(sk, opt, opt->tot_len);
920 inet_csk(newsk)->icsk_ext_hdr_len = 0;
922 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
923 newnp->opt->opt_flen);
925 dccp_sync_mss(newsk, dst_mtu(dst));
927 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
929 __inet6_hash(&dccp_hashinfo, newsk);
930 inet_inherit_port(&dccp_hashinfo, sk, newsk);
935 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
937 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
938 if (opt && opt != np->opt)
939 sock_kfree_s(sk, opt, opt->tot_len);
944 /* The socket must have it's spinlock held when we get
947 * We have a potential double-lock case here, so even when
948 * doing backlog processing we use the BH locking scheme.
949 * This is because we cannot sleep with the original spinlock
952 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
954 struct ipv6_pinfo *np = inet6_sk(sk);
955 struct sk_buff *opt_skb = NULL;
957 /* Imagine: socket is IPv6. IPv4 packet arrives,
958 goes to IPv4 receive handler and backlogged.
959 From backlog it always goes here. Kerboom...
960 Fortunately, dccp_rcv_established and rcv_established
961 handle them correctly, but it is not case with
962 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
965 if (skb->protocol == htons(ETH_P_IP))
966 return dccp_v4_do_rcv(sk, skb);
968 if (sk_filter(sk, skb, 0))
972 * socket locking is here for SMP purposes as backlog rcv
973 * is currently called with bh processing disabled.
976 /* Do Stevens' IPV6_PKTOPTIONS.
978 Yes, guys, it is the only place in our code, where we
979 may make it not affecting IPv4.
980 The rest of code is protocol independent,
981 and I do not like idea to uglify IPv4.
983 Actually, all the idea behind IPV6_PKTOPTIONS
984 looks not very well thought. For now we latch
985 options, received in the last packet, enqueued
986 by tcp. Feel free to propose better solution.
990 opt_skb = skb_clone(skb, GFP_ATOMIC);
992 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
993 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
998 if (sk->sk_state == DCCP_LISTEN) {
999 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
1004 * Queue it on the new socket if the new socket is active,
1005 * otherwise we just shortcircuit this and continue with
1009 if (dccp_child_process(sk, nsk, skb))
1012 __kfree_skb(opt_skb);
1017 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
1022 dccp_v6_ctl_send_reset(skb);
1025 __kfree_skb(opt_skb);
1030 static int dccp_v6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
1032 const struct dccp_hdr *dh;
1033 struct sk_buff *skb = *pskb;
1037 /* Step 1: Check header basics: */
1039 if (dccp_invalid_packet(skb))
1044 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
1045 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
1047 if (dccp_packet_without_ack(skb))
1048 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
1050 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
1053 * Look up flow ID in table and get corresponding socket */
1054 sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
1056 &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
1061 * Generate Reset(No Connection) unless P.type == Reset
1062 * Drop packet and return
1065 goto no_dccp_socket;
1069 * ... or S.state == TIMEWAIT,
1070 * Generate Reset(No Connection) unless P.type == Reset
1071 * Drop packet and return
1074 if (sk->sk_state == DCCP_TIME_WAIT)
1077 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1078 goto discard_and_relse;
1080 if (sk_filter(sk, skb, 0))
1081 goto discard_and_relse;
1087 if (!sock_owned_by_user(sk))
1088 rc = dccp_v6_do_rcv(sk, skb);
1090 sk_add_backlog(sk, skb);
1097 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1101 * Generate Reset(No Connection) unless P.type == Reset
1102 * Drop packet and return
1104 if (dh->dccph_type != DCCP_PKT_RESET) {
1105 DCCP_SKB_CB(skb)->dccpd_reset_code =
1106 DCCP_RESET_CODE_NO_CONNECTION;
1107 dccp_v6_ctl_send_reset(skb);
1123 inet_twsk_put((struct inet_timewait_sock *)sk);
1124 goto no_dccp_socket;
1127 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1128 .queue_xmit = inet6_csk_xmit,
1129 .send_check = dccp_v6_send_check,
1130 .rebuild_header = inet6_sk_rebuild_header,
1131 .conn_request = dccp_v6_conn_request,
1132 .syn_recv_sock = dccp_v6_request_recv_sock,
1133 .net_header_len = sizeof(struct ipv6hdr),
1134 .setsockopt = ipv6_setsockopt,
1135 .getsockopt = ipv6_getsockopt,
1136 .addr2sockaddr = inet6_csk_addr2sockaddr,
1137 .sockaddr_len = sizeof(struct sockaddr_in6)
1141 * DCCP over IPv4 via INET6 API
1143 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1144 .queue_xmit = ip_queue_xmit,
1145 .send_check = dccp_v4_send_check,
1146 .rebuild_header = inet_sk_rebuild_header,
1147 .conn_request = dccp_v6_conn_request,
1148 .syn_recv_sock = dccp_v6_request_recv_sock,
1149 .net_header_len = sizeof(struct iphdr),
1150 .setsockopt = ipv6_setsockopt,
1151 .getsockopt = ipv6_getsockopt,
1152 .addr2sockaddr = inet6_csk_addr2sockaddr,
1153 .sockaddr_len = sizeof(struct sockaddr_in6)
1156 /* NOTE: A lot of things set to zero explicitly by call to
1157 * sk_alloc() so need not be done here.
1159 static int dccp_v6_init_sock(struct sock *sk)
1161 int err = dccp_v4_init_sock(sk);
1164 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1169 static int dccp_v6_destroy_sock(struct sock *sk)
1171 dccp_v4_destroy_sock(sk);
1172 return inet6_destroy_sock(sk);
1175 static struct proto dccp_v6_prot = {
1177 .owner = THIS_MODULE,
1178 .close = dccp_close,
1179 .connect = dccp_v6_connect,
1180 .disconnect = dccp_disconnect,
1181 .ioctl = dccp_ioctl,
1182 .init = dccp_v6_init_sock,
1183 .setsockopt = dccp_setsockopt,
1184 .getsockopt = dccp_getsockopt,
1185 .sendmsg = dccp_sendmsg,
1186 .recvmsg = dccp_recvmsg,
1187 .backlog_rcv = dccp_v6_do_rcv,
1188 .hash = dccp_v6_hash,
1189 .unhash = dccp_unhash,
1190 .accept = inet_csk_accept,
1191 .get_port = dccp_v6_get_port,
1192 .shutdown = dccp_shutdown,
1193 .destroy = dccp_v6_destroy_sock,
1194 .orphan_count = &dccp_orphan_count,
1195 .max_header = MAX_DCCP_HEADER,
1196 .obj_size = sizeof(struct dccp6_sock),
1197 .rsk_prot = &dccp6_request_sock_ops,
1198 .twsk_prot = &dccp6_timewait_sock_ops,
1201 static struct inet6_protocol dccp_v6_protocol = {
1202 .handler = dccp_v6_rcv,
1203 .err_handler = dccp_v6_err,
1204 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1207 static struct proto_ops inet6_dccp_ops = {
1209 .owner = THIS_MODULE,
1210 .release = inet6_release,
1212 .connect = inet_stream_connect,
1213 .socketpair = sock_no_socketpair,
1214 .accept = inet_accept,
1215 .getname = inet6_getname,
1217 .ioctl = inet6_ioctl,
1218 .listen = inet_dccp_listen,
1219 .shutdown = inet_shutdown,
1220 .setsockopt = sock_common_setsockopt,
1221 .getsockopt = sock_common_getsockopt,
1222 .sendmsg = inet_sendmsg,
1223 .recvmsg = sock_common_recvmsg,
1224 .mmap = sock_no_mmap,
1225 .sendpage = sock_no_sendpage,
1228 static struct inet_protosw dccp_v6_protosw = {
1230 .protocol = IPPROTO_DCCP,
1231 .prot = &dccp_v6_prot,
1232 .ops = &inet6_dccp_ops,
1234 .flags = INET_PROTOSW_ICSK,
1237 static int __init dccp_v6_init(void)
1239 int err = proto_register(&dccp_v6_prot, 1);
1244 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1246 goto out_unregister_proto;
1248 inet6_register_protosw(&dccp_v6_protosw);
1251 out_unregister_proto:
1252 proto_unregister(&dccp_v6_prot);
1256 static void __exit dccp_v6_exit(void)
1258 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1259 inet6_unregister_protosw(&dccp_v6_protosw);
1260 proto_unregister(&dccp_v6_prot);
1263 module_init(dccp_v6_init);
1264 module_exit(dccp_v6_exit);
1267 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1268 * values directly, Also cover the case where the protocol is not specified,
1269 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1271 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
1272 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
1273 MODULE_LICENSE("GPL");
1274 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1275 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");