remove the build INITRAMFS
[qi-kernel.git] / net / dccp / ipv6.c
blobe48ca5d45658468176bb60c8cbb53a587feea72d
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 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
38 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
39 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
41 static void dccp_v6_hash(struct sock *sk)
43 if (sk->sk_state != DCCP_CLOSED) {
44 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
45 inet_hash(sk);
46 return;
48 local_bh_disable();
49 __inet6_hash(sk);
50 local_bh_enable();
54 /* add pseudo-header to DCCP checksum stored in skb->csum */
55 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
56 struct in6_addr *saddr,
57 struct in6_addr *daddr)
59 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
62 static inline void dccp_v6_send_check(struct sock *sk, int unused_value,
63 struct sk_buff *skb)
65 struct ipv6_pinfo *np = inet6_sk(sk);
66 struct dccp_hdr *dh = dccp_hdr(skb);
68 dccp_csum_outgoing(skb);
69 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr);
72 static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
73 __be16 sport, __be16 dport )
75 return secure_tcpv6_sequence_number(saddr, daddr, sport, dport);
78 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb)
80 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
81 ipv6_hdr(skb)->saddr.s6_addr32,
82 dccp_hdr(skb)->dccph_dport,
83 dccp_hdr(skb)->dccph_sport );
87 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
88 u8 type, u8 code, int offset, __be32 info)
90 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
91 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
92 struct dccp_sock *dp;
93 struct ipv6_pinfo *np;
94 struct sock *sk;
95 int err;
96 __u64 seq;
97 struct net *net = dev_net(skb->dev);
99 if (skb->len < offset + sizeof(*dh) ||
100 skb->len < offset + __dccp_basic_hdr_len(dh)) {
101 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
102 ICMP6_MIB_INERRORS);
103 return;
106 sk = inet6_lookup(net, &dccp_hashinfo,
107 &hdr->daddr, dh->dccph_dport,
108 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
110 if (sk == NULL) {
111 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
112 ICMP6_MIB_INERRORS);
113 return;
116 if (sk->sk_state == DCCP_TIME_WAIT) {
117 inet_twsk_put(inet_twsk(sk));
118 return;
121 bh_lock_sock(sk);
122 if (sock_owned_by_user(sk))
123 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
125 if (sk->sk_state == DCCP_CLOSED)
126 goto out;
128 dp = dccp_sk(sk);
129 seq = dccp_hdr_seq(dh);
130 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
131 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
132 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
133 goto out;
136 np = inet6_sk(sk);
138 if (type == ICMPV6_PKT_TOOBIG) {
139 struct dst_entry *dst = NULL;
141 if (sock_owned_by_user(sk))
142 goto out;
143 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
144 goto out;
146 /* icmp should have updated the destination cache entry */
147 dst = __sk_dst_check(sk, np->dst_cookie);
148 if (dst == NULL) {
149 struct inet_sock *inet = inet_sk(sk);
150 struct flowi fl;
152 /* BUGGG_FUTURE: Again, it is not clear how
153 to handle rthdr case. Ignore this complexity
154 for now.
156 memset(&fl, 0, sizeof(fl));
157 fl.proto = IPPROTO_DCCP;
158 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
159 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
160 fl.oif = sk->sk_bound_dev_if;
161 fl.fl_ip_dport = inet->dport;
162 fl.fl_ip_sport = inet->sport;
163 security_sk_classify_flow(sk, &fl);
165 err = ip6_dst_lookup(sk, &dst, &fl);
166 if (err) {
167 sk->sk_err_soft = -err;
168 goto out;
171 err = xfrm_lookup(net, &dst, &fl, sk, 0);
172 if (err < 0) {
173 sk->sk_err_soft = -err;
174 goto out;
176 } else
177 dst_hold(dst);
179 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
180 dccp_sync_mss(sk, dst_mtu(dst));
181 } /* else let the usual retransmit timer handle it */
182 dst_release(dst);
183 goto out;
186 icmpv6_err_convert(type, code, &err);
188 /* Might be for an request_sock */
189 switch (sk->sk_state) {
190 struct request_sock *req, **prev;
191 case DCCP_LISTEN:
192 if (sock_owned_by_user(sk))
193 goto out;
195 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
196 &hdr->daddr, &hdr->saddr,
197 inet6_iif(skb));
198 if (req == NULL)
199 goto out;
202 * ICMPs are not backlogged, hence we cannot get an established
203 * socket here.
205 WARN_ON(req->sk != NULL);
207 if (seq != dccp_rsk(req)->dreq_iss) {
208 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
209 goto out;
212 inet_csk_reqsk_queue_drop(sk, req, prev);
213 goto out;
215 case DCCP_REQUESTING:
216 case DCCP_RESPOND: /* Cannot happen.
217 It can, it SYNs are crossed. --ANK */
218 if (!sock_owned_by_user(sk)) {
219 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
220 sk->sk_err = err;
222 * Wake people up to see the error
223 * (see connect in sock.c)
225 sk->sk_error_report(sk);
226 dccp_done(sk);
227 } else
228 sk->sk_err_soft = err;
229 goto out;
232 if (!sock_owned_by_user(sk) && np->recverr) {
233 sk->sk_err = err;
234 sk->sk_error_report(sk);
235 } else
236 sk->sk_err_soft = err;
238 out:
239 bh_unlock_sock(sk);
240 sock_put(sk);
244 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req)
246 struct inet6_request_sock *ireq6 = inet6_rsk(req);
247 struct ipv6_pinfo *np = inet6_sk(sk);
248 struct sk_buff *skb;
249 struct ipv6_txoptions *opt = NULL;
250 struct in6_addr *final_p = NULL, final;
251 struct flowi fl;
252 int err = -1;
253 struct dst_entry *dst;
255 memset(&fl, 0, sizeof(fl));
256 fl.proto = IPPROTO_DCCP;
257 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
258 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
259 fl.fl6_flowlabel = 0;
260 fl.oif = ireq6->iif;
261 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
262 fl.fl_ip_sport = inet_rsk(req)->loc_port;
263 security_req_classify_flow(req, &fl);
265 opt = np->opt;
267 if (opt != NULL && opt->srcrt != NULL) {
268 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
270 ipv6_addr_copy(&final, &fl.fl6_dst);
271 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
272 final_p = &final;
275 err = ip6_dst_lookup(sk, &dst, &fl);
276 if (err)
277 goto done;
279 if (final_p)
280 ipv6_addr_copy(&fl.fl6_dst, final_p);
282 err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0);
283 if (err < 0)
284 goto done;
286 skb = dccp_make_response(sk, dst, req);
287 if (skb != NULL) {
288 struct dccp_hdr *dh = dccp_hdr(skb);
290 dh->dccph_checksum = dccp_v6_csum_finish(skb,
291 &ireq6->loc_addr,
292 &ireq6->rmt_addr);
293 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
294 err = ip6_xmit(sk, skb, &fl, opt, 0);
295 err = net_xmit_eval(err);
298 done:
299 if (opt != NULL && opt != np->opt)
300 sock_kfree_s(sk, opt, opt->tot_len);
301 dst_release(dst);
302 return err;
305 static void dccp_v6_reqsk_destructor(struct request_sock *req)
307 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
308 if (inet6_rsk(req)->pktopts != NULL)
309 kfree_skb(inet6_rsk(req)->pktopts);
312 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
314 struct ipv6hdr *rxip6h;
315 struct sk_buff *skb;
316 struct flowi fl;
317 struct net *net = dev_net(skb_dst(rxskb)->dev);
318 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
319 struct dst_entry *dst;
321 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
322 return;
324 if (!ipv6_unicast_destination(rxskb))
325 return;
327 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
328 if (skb == NULL)
329 return;
331 rxip6h = ipv6_hdr(rxskb);
332 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
333 &rxip6h->daddr);
335 memset(&fl, 0, sizeof(fl));
336 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
337 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
339 fl.proto = IPPROTO_DCCP;
340 fl.oif = inet6_iif(rxskb);
341 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
342 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
343 security_skb_classify_flow(rxskb, &fl);
345 /* sk = NULL, but it is safe for now. RST socket required. */
346 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) {
347 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) {
348 skb_dst_set(skb, dst);
349 ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
350 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
351 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
352 return;
356 kfree_skb(skb);
359 static struct request_sock_ops dccp6_request_sock_ops = {
360 .family = AF_INET6,
361 .obj_size = sizeof(struct dccp6_request_sock),
362 .rtx_syn_ack = dccp_v6_send_response,
363 .send_ack = dccp_reqsk_send_ack,
364 .destructor = dccp_v6_reqsk_destructor,
365 .send_reset = dccp_v6_ctl_send_reset,
368 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
370 const struct dccp_hdr *dh = dccp_hdr(skb);
371 const struct ipv6hdr *iph = ipv6_hdr(skb);
372 struct sock *nsk;
373 struct request_sock **prev;
374 /* Find possible connection requests. */
375 struct request_sock *req = inet6_csk_search_req(sk, &prev,
376 dh->dccph_sport,
377 &iph->saddr,
378 &iph->daddr,
379 inet6_iif(skb));
380 if (req != NULL)
381 return dccp_check_req(sk, skb, req, prev);
383 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
384 &iph->saddr, dh->dccph_sport,
385 &iph->daddr, ntohs(dh->dccph_dport),
386 inet6_iif(skb));
387 if (nsk != NULL) {
388 if (nsk->sk_state != DCCP_TIME_WAIT) {
389 bh_lock_sock(nsk);
390 return nsk;
392 inet_twsk_put(inet_twsk(nsk));
393 return NULL;
396 return sk;
399 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
401 struct request_sock *req;
402 struct dccp_request_sock *dreq;
403 struct inet6_request_sock *ireq6;
404 struct ipv6_pinfo *np = inet6_sk(sk);
405 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
406 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
408 if (skb->protocol == htons(ETH_P_IP))
409 return dccp_v4_conn_request(sk, skb);
411 if (!ipv6_unicast_destination(skb))
412 return 0; /* discard, don't send a reset here */
414 if (dccp_bad_service_code(sk, service)) {
415 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
416 goto drop;
419 * There are no SYN attacks on IPv6, yet...
421 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
422 if (inet_csk_reqsk_queue_is_full(sk))
423 goto drop;
425 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
426 goto drop;
428 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
429 if (req == NULL)
430 goto drop;
432 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
433 goto drop_and_free;
435 dreq = dccp_rsk(req);
436 if (dccp_parse_options(sk, dreq, skb))
437 goto drop_and_free;
439 if (security_inet_conn_request(sk, skb, req))
440 goto drop_and_free;
442 ireq6 = inet6_rsk(req);
443 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
444 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
446 if (ipv6_opt_accepted(sk, skb) ||
447 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
448 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
449 atomic_inc(&skb->users);
450 ireq6->pktopts = skb;
452 ireq6->iif = sk->sk_bound_dev_if;
454 /* So that link locals have meaning */
455 if (!sk->sk_bound_dev_if &&
456 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
457 ireq6->iif = inet6_iif(skb);
460 * Step 3: Process LISTEN state
462 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
464 * In fact we defer setting S.GSR, S.SWL, S.SWH to
465 * dccp_create_openreq_child.
467 dreq->dreq_isr = dcb->dccpd_seq;
468 dreq->dreq_iss = dccp_v6_init_sequence(skb);
469 dreq->dreq_service = service;
471 if (dccp_v6_send_response(sk, req))
472 goto drop_and_free;
474 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
475 return 0;
477 drop_and_free:
478 reqsk_free(req);
479 drop:
480 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
481 return -1;
484 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
485 struct sk_buff *skb,
486 struct request_sock *req,
487 struct dst_entry *dst)
489 struct inet6_request_sock *ireq6 = inet6_rsk(req);
490 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
491 struct inet_sock *newinet;
492 struct dccp_sock *newdp;
493 struct dccp6_sock *newdp6;
494 struct sock *newsk;
495 struct ipv6_txoptions *opt;
497 if (skb->protocol == htons(ETH_P_IP)) {
499 * v6 mapped
501 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
502 if (newsk == NULL)
503 return NULL;
505 newdp6 = (struct dccp6_sock *)newsk;
506 newdp = dccp_sk(newsk);
507 newinet = inet_sk(newsk);
508 newinet->pinet6 = &newdp6->inet6;
509 newnp = inet6_sk(newsk);
511 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
513 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
514 newinet->daddr);
516 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
517 newinet->saddr);
519 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
521 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
522 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
523 newnp->pktoptions = NULL;
524 newnp->opt = NULL;
525 newnp->mcast_oif = inet6_iif(skb);
526 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
529 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
530 * here, dccp_create_openreq_child now does this for us, see the comment in
531 * that function for the gory details. -acme
534 /* It is tricky place. Until this moment IPv4 tcp
535 worked with IPv6 icsk.icsk_af_ops.
536 Sync it now.
538 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
540 return newsk;
543 opt = np->opt;
545 if (sk_acceptq_is_full(sk))
546 goto out_overflow;
548 if (dst == NULL) {
549 struct in6_addr *final_p = NULL, final;
550 struct flowi fl;
552 memset(&fl, 0, sizeof(fl));
553 fl.proto = IPPROTO_DCCP;
554 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
555 if (opt != NULL && opt->srcrt != NULL) {
556 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
558 ipv6_addr_copy(&final, &fl.fl6_dst);
559 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
560 final_p = &final;
562 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
563 fl.oif = sk->sk_bound_dev_if;
564 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
565 fl.fl_ip_sport = inet_rsk(req)->loc_port;
566 security_sk_classify_flow(sk, &fl);
568 if (ip6_dst_lookup(sk, &dst, &fl))
569 goto out;
571 if (final_p)
572 ipv6_addr_copy(&fl.fl6_dst, final_p);
574 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
575 goto out;
578 newsk = dccp_create_openreq_child(sk, req, skb);
579 if (newsk == NULL)
580 goto out;
583 * No need to charge this sock to the relevant IPv6 refcnt debug socks
584 * count here, dccp_create_openreq_child now does this for us, see the
585 * comment in that function for the gory details. -acme
588 __ip6_dst_store(newsk, dst, NULL, NULL);
589 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
590 NETIF_F_TSO);
591 newdp6 = (struct dccp6_sock *)newsk;
592 newinet = inet_sk(newsk);
593 newinet->pinet6 = &newdp6->inet6;
594 newdp = dccp_sk(newsk);
595 newnp = inet6_sk(newsk);
597 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
599 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
600 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
601 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
602 newsk->sk_bound_dev_if = ireq6->iif;
604 /* Now IPv6 options...
606 First: no IPv4 options.
608 newinet->opt = NULL;
610 /* Clone RX bits */
611 newnp->rxopt.all = np->rxopt.all;
613 /* Clone pktoptions received with SYN */
614 newnp->pktoptions = NULL;
615 if (ireq6->pktopts != NULL) {
616 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
617 kfree_skb(ireq6->pktopts);
618 ireq6->pktopts = NULL;
619 if (newnp->pktoptions)
620 skb_set_owner_r(newnp->pktoptions, newsk);
622 newnp->opt = NULL;
623 newnp->mcast_oif = inet6_iif(skb);
624 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
627 * Clone native IPv6 options from listening socket (if any)
629 * Yes, keeping reference count would be much more clever, but we make
630 * one more one thing there: reattach optmem to newsk.
632 if (opt != NULL) {
633 newnp->opt = ipv6_dup_options(newsk, opt);
634 if (opt != np->opt)
635 sock_kfree_s(sk, opt, opt->tot_len);
638 inet_csk(newsk)->icsk_ext_hdr_len = 0;
639 if (newnp->opt != NULL)
640 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
641 newnp->opt->opt_flen);
643 dccp_sync_mss(newsk, dst_mtu(dst));
645 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
647 __inet6_hash(newsk);
648 __inet_inherit_port(sk, newsk);
650 return newsk;
652 out_overflow:
653 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
654 out:
655 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
656 if (opt != NULL && opt != np->opt)
657 sock_kfree_s(sk, opt, opt->tot_len);
658 dst_release(dst);
659 return NULL;
662 /* The socket must have it's spinlock held when we get
663 * here.
665 * We have a potential double-lock case here, so even when
666 * doing backlog processing we use the BH locking scheme.
667 * This is because we cannot sleep with the original spinlock
668 * held.
670 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
672 struct ipv6_pinfo *np = inet6_sk(sk);
673 struct sk_buff *opt_skb = NULL;
675 /* Imagine: socket is IPv6. IPv4 packet arrives,
676 goes to IPv4 receive handler and backlogged.
677 From backlog it always goes here. Kerboom...
678 Fortunately, dccp_rcv_established and rcv_established
679 handle them correctly, but it is not case with
680 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
683 if (skb->protocol == htons(ETH_P_IP))
684 return dccp_v4_do_rcv(sk, skb);
686 if (sk_filter(sk, skb))
687 goto discard;
690 * socket locking is here for SMP purposes as backlog rcv is currently
691 * called with bh processing disabled.
694 /* Do Stevens' IPV6_PKTOPTIONS.
696 Yes, guys, it is the only place in our code, where we
697 may make it not affecting IPv4.
698 The rest of code is protocol independent,
699 and I do not like idea to uglify IPv4.
701 Actually, all the idea behind IPV6_PKTOPTIONS
702 looks not very well thought. For now we latch
703 options, received in the last packet, enqueued
704 by tcp. Feel free to propose better solution.
705 --ANK (980728)
707 if (np->rxopt.all)
709 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
710 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
712 opt_skb = skb_clone(skb, GFP_ATOMIC);
714 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
715 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
716 goto reset;
717 if (opt_skb) {
718 /* XXX This is where we would goto ipv6_pktoptions. */
719 __kfree_skb(opt_skb);
721 return 0;
725 * Step 3: Process LISTEN state
726 * If S.state == LISTEN,
727 * If P.type == Request or P contains a valid Init Cookie option,
728 * (* Must scan the packet's options to check for Init
729 * Cookies. Only Init Cookies are processed here,
730 * however; other options are processed in Step 8. This
731 * scan need only be performed if the endpoint uses Init
732 * Cookies *)
733 * (* Generate a new socket and switch to that socket *)
734 * Set S := new socket for this port pair
735 * S.state = RESPOND
736 * Choose S.ISS (initial seqno) or set from Init Cookies
737 * Initialize S.GAR := S.ISS
738 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
739 * Continue with S.state == RESPOND
740 * (* A Response packet will be generated in Step 11 *)
741 * Otherwise,
742 * Generate Reset(No Connection) unless P.type == Reset
743 * Drop packet and return
745 * NOTE: the check for the packet types is done in
746 * dccp_rcv_state_process
748 if (sk->sk_state == DCCP_LISTEN) {
749 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
751 if (nsk == NULL)
752 goto discard;
754 * Queue it on the new socket if the new socket is active,
755 * otherwise we just shortcircuit this and continue with
756 * the new socket..
758 if (nsk != sk) {
759 if (dccp_child_process(sk, nsk, skb))
760 goto reset;
761 if (opt_skb != NULL)
762 __kfree_skb(opt_skb);
763 return 0;
767 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
768 goto reset;
769 if (opt_skb) {
770 /* XXX This is where we would goto ipv6_pktoptions. */
771 __kfree_skb(opt_skb);
773 return 0;
775 reset:
776 dccp_v6_ctl_send_reset(sk, skb);
777 discard:
778 if (opt_skb != NULL)
779 __kfree_skb(opt_skb);
780 kfree_skb(skb);
781 return 0;
784 static int dccp_v6_rcv(struct sk_buff *skb)
786 const struct dccp_hdr *dh;
787 struct sock *sk;
788 int min_cov;
790 /* Step 1: Check header basics */
792 if (dccp_invalid_packet(skb))
793 goto discard_it;
795 /* Step 1: If header checksum is incorrect, drop packet and return. */
796 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
797 &ipv6_hdr(skb)->daddr)) {
798 DCCP_WARN("dropped packet with invalid checksum\n");
799 goto discard_it;
802 dh = dccp_hdr(skb);
804 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
805 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
807 if (dccp_packet_without_ack(skb))
808 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
809 else
810 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
812 /* Step 2:
813 * Look up flow ID in table and get corresponding socket */
814 sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
815 dh->dccph_sport, dh->dccph_dport);
817 * Step 2:
818 * If no socket ...
820 if (sk == NULL) {
821 dccp_pr_debug("failed to look up flow ID in table and "
822 "get corresponding socket\n");
823 goto no_dccp_socket;
827 * Step 2:
828 * ... or S.state == TIMEWAIT,
829 * Generate Reset(No Connection) unless P.type == Reset
830 * Drop packet and return
832 if (sk->sk_state == DCCP_TIME_WAIT) {
833 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
834 inet_twsk_put(inet_twsk(sk));
835 goto no_dccp_socket;
839 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
840 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
841 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
843 min_cov = dccp_sk(sk)->dccps_pcrlen;
844 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
845 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
846 dh->dccph_cscov, min_cov);
847 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
848 goto discard_and_relse;
851 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
852 goto discard_and_relse;
854 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
856 no_dccp_socket:
857 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
858 goto discard_it;
860 * Step 2:
861 * If no socket ...
862 * Generate Reset(No Connection) unless P.type == Reset
863 * Drop packet and return
865 if (dh->dccph_type != DCCP_PKT_RESET) {
866 DCCP_SKB_CB(skb)->dccpd_reset_code =
867 DCCP_RESET_CODE_NO_CONNECTION;
868 dccp_v6_ctl_send_reset(sk, skb);
871 discard_it:
872 kfree_skb(skb);
873 return 0;
875 discard_and_relse:
876 sock_put(sk);
877 goto discard_it;
880 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
881 int addr_len)
883 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
884 struct inet_connection_sock *icsk = inet_csk(sk);
885 struct inet_sock *inet = inet_sk(sk);
886 struct ipv6_pinfo *np = inet6_sk(sk);
887 struct dccp_sock *dp = dccp_sk(sk);
888 struct in6_addr *saddr = NULL, *final_p = NULL, final;
889 struct flowi fl;
890 struct dst_entry *dst;
891 int addr_type;
892 int err;
894 dp->dccps_role = DCCP_ROLE_CLIENT;
896 if (addr_len < SIN6_LEN_RFC2133)
897 return -EINVAL;
899 if (usin->sin6_family != AF_INET6)
900 return -EAFNOSUPPORT;
902 memset(&fl, 0, sizeof(fl));
904 if (np->sndflow) {
905 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
906 IP6_ECN_flow_init(fl.fl6_flowlabel);
907 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
908 struct ip6_flowlabel *flowlabel;
909 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
910 if (flowlabel == NULL)
911 return -EINVAL;
912 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
913 fl6_sock_release(flowlabel);
917 * connect() to INADDR_ANY means loopback (BSD'ism).
919 if (ipv6_addr_any(&usin->sin6_addr))
920 usin->sin6_addr.s6_addr[15] = 1;
922 addr_type = ipv6_addr_type(&usin->sin6_addr);
924 if (addr_type & IPV6_ADDR_MULTICAST)
925 return -ENETUNREACH;
927 if (addr_type & IPV6_ADDR_LINKLOCAL) {
928 if (addr_len >= sizeof(struct sockaddr_in6) &&
929 usin->sin6_scope_id) {
930 /* If interface is set while binding, indices
931 * must coincide.
933 if (sk->sk_bound_dev_if &&
934 sk->sk_bound_dev_if != usin->sin6_scope_id)
935 return -EINVAL;
937 sk->sk_bound_dev_if = usin->sin6_scope_id;
940 /* Connect to link-local address requires an interface */
941 if (!sk->sk_bound_dev_if)
942 return -EINVAL;
945 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
946 np->flow_label = fl.fl6_flowlabel;
949 * DCCP over IPv4
951 if (addr_type == IPV6_ADDR_MAPPED) {
952 u32 exthdrlen = icsk->icsk_ext_hdr_len;
953 struct sockaddr_in sin;
955 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
957 if (__ipv6_only_sock(sk))
958 return -ENETUNREACH;
960 sin.sin_family = AF_INET;
961 sin.sin_port = usin->sin6_port;
962 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
964 icsk->icsk_af_ops = &dccp_ipv6_mapped;
965 sk->sk_backlog_rcv = dccp_v4_do_rcv;
967 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
968 if (err) {
969 icsk->icsk_ext_hdr_len = exthdrlen;
970 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
971 sk->sk_backlog_rcv = dccp_v6_do_rcv;
972 goto failure;
973 } else {
974 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
975 inet->saddr);
976 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
977 inet->rcv_saddr);
980 return err;
983 if (!ipv6_addr_any(&np->rcv_saddr))
984 saddr = &np->rcv_saddr;
986 fl.proto = IPPROTO_DCCP;
987 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
988 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
989 fl.oif = sk->sk_bound_dev_if;
990 fl.fl_ip_dport = usin->sin6_port;
991 fl.fl_ip_sport = inet->sport;
992 security_sk_classify_flow(sk, &fl);
994 if (np->opt != NULL && np->opt->srcrt != NULL) {
995 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
997 ipv6_addr_copy(&final, &fl.fl6_dst);
998 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
999 final_p = &final;
1002 err = ip6_dst_lookup(sk, &dst, &fl);
1003 if (err)
1004 goto failure;
1006 if (final_p)
1007 ipv6_addr_copy(&fl.fl6_dst, final_p);
1009 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
1010 if (err < 0) {
1011 if (err == -EREMOTE)
1012 err = ip6_dst_blackhole(sk, &dst, &fl);
1013 if (err < 0)
1014 goto failure;
1017 if (saddr == NULL) {
1018 saddr = &fl.fl6_src;
1019 ipv6_addr_copy(&np->rcv_saddr, saddr);
1022 /* set the source address */
1023 ipv6_addr_copy(&np->saddr, saddr);
1024 inet->rcv_saddr = LOOPBACK4_IPV6;
1026 __ip6_dst_store(sk, dst, NULL, NULL);
1028 icsk->icsk_ext_hdr_len = 0;
1029 if (np->opt != NULL)
1030 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1031 np->opt->opt_nflen);
1033 inet->dport = usin->sin6_port;
1035 dccp_set_state(sk, DCCP_REQUESTING);
1036 err = inet6_hash_connect(&dccp_death_row, sk);
1037 if (err)
1038 goto late_failure;
1040 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1041 np->daddr.s6_addr32,
1042 inet->sport, inet->dport);
1043 err = dccp_connect(sk);
1044 if (err)
1045 goto late_failure;
1047 return 0;
1049 late_failure:
1050 dccp_set_state(sk, DCCP_CLOSED);
1051 __sk_dst_reset(sk);
1052 failure:
1053 inet->dport = 0;
1054 sk->sk_route_caps = 0;
1055 return err;
1058 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1059 .queue_xmit = inet6_csk_xmit,
1060 .send_check = dccp_v6_send_check,
1061 .rebuild_header = inet6_sk_rebuild_header,
1062 .conn_request = dccp_v6_conn_request,
1063 .syn_recv_sock = dccp_v6_request_recv_sock,
1064 .net_header_len = sizeof(struct ipv6hdr),
1065 .setsockopt = ipv6_setsockopt,
1066 .getsockopt = ipv6_getsockopt,
1067 .addr2sockaddr = inet6_csk_addr2sockaddr,
1068 .sockaddr_len = sizeof(struct sockaddr_in6),
1069 .bind_conflict = inet6_csk_bind_conflict,
1070 #ifdef CONFIG_COMPAT
1071 .compat_setsockopt = compat_ipv6_setsockopt,
1072 .compat_getsockopt = compat_ipv6_getsockopt,
1073 #endif
1077 * DCCP over IPv4 via INET6 API
1079 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1080 .queue_xmit = ip_queue_xmit,
1081 .send_check = dccp_v4_send_check,
1082 .rebuild_header = inet_sk_rebuild_header,
1083 .conn_request = dccp_v6_conn_request,
1084 .syn_recv_sock = dccp_v6_request_recv_sock,
1085 .net_header_len = sizeof(struct iphdr),
1086 .setsockopt = ipv6_setsockopt,
1087 .getsockopt = ipv6_getsockopt,
1088 .addr2sockaddr = inet6_csk_addr2sockaddr,
1089 .sockaddr_len = sizeof(struct sockaddr_in6),
1090 #ifdef CONFIG_COMPAT
1091 .compat_setsockopt = compat_ipv6_setsockopt,
1092 .compat_getsockopt = compat_ipv6_getsockopt,
1093 #endif
1096 /* NOTE: A lot of things set to zero explicitly by call to
1097 * sk_alloc() so need not be done here.
1099 static int dccp_v6_init_sock(struct sock *sk)
1101 static __u8 dccp_v6_ctl_sock_initialized;
1102 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1104 if (err == 0) {
1105 if (unlikely(!dccp_v6_ctl_sock_initialized))
1106 dccp_v6_ctl_sock_initialized = 1;
1107 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1110 return err;
1113 static void dccp_v6_destroy_sock(struct sock *sk)
1115 dccp_destroy_sock(sk);
1116 inet6_destroy_sock(sk);
1119 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1120 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1123 static struct proto dccp_v6_prot = {
1124 .name = "DCCPv6",
1125 .owner = THIS_MODULE,
1126 .close = dccp_close,
1127 .connect = dccp_v6_connect,
1128 .disconnect = dccp_disconnect,
1129 .ioctl = dccp_ioctl,
1130 .init = dccp_v6_init_sock,
1131 .setsockopt = dccp_setsockopt,
1132 .getsockopt = dccp_getsockopt,
1133 .sendmsg = dccp_sendmsg,
1134 .recvmsg = dccp_recvmsg,
1135 .backlog_rcv = dccp_v6_do_rcv,
1136 .hash = dccp_v6_hash,
1137 .unhash = inet_unhash,
1138 .accept = inet_csk_accept,
1139 .get_port = inet_csk_get_port,
1140 .shutdown = dccp_shutdown,
1141 .destroy = dccp_v6_destroy_sock,
1142 .orphan_count = &dccp_orphan_count,
1143 .max_header = MAX_DCCP_HEADER,
1144 .obj_size = sizeof(struct dccp6_sock),
1145 .slab_flags = SLAB_DESTROY_BY_RCU,
1146 .rsk_prot = &dccp6_request_sock_ops,
1147 .twsk_prot = &dccp6_timewait_sock_ops,
1148 .h.hashinfo = &dccp_hashinfo,
1149 #ifdef CONFIG_COMPAT
1150 .compat_setsockopt = compat_dccp_setsockopt,
1151 .compat_getsockopt = compat_dccp_getsockopt,
1152 #endif
1155 static const struct inet6_protocol dccp_v6_protocol = {
1156 .handler = dccp_v6_rcv,
1157 .err_handler = dccp_v6_err,
1158 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1161 static const struct proto_ops inet6_dccp_ops = {
1162 .family = PF_INET6,
1163 .owner = THIS_MODULE,
1164 .release = inet6_release,
1165 .bind = inet6_bind,
1166 .connect = inet_stream_connect,
1167 .socketpair = sock_no_socketpair,
1168 .accept = inet_accept,
1169 .getname = inet6_getname,
1170 .poll = dccp_poll,
1171 .ioctl = inet6_ioctl,
1172 .listen = inet_dccp_listen,
1173 .shutdown = inet_shutdown,
1174 .setsockopt = sock_common_setsockopt,
1175 .getsockopt = sock_common_getsockopt,
1176 .sendmsg = inet_sendmsg,
1177 .recvmsg = sock_common_recvmsg,
1178 .mmap = sock_no_mmap,
1179 .sendpage = sock_no_sendpage,
1180 #ifdef CONFIG_COMPAT
1181 .compat_setsockopt = compat_sock_common_setsockopt,
1182 .compat_getsockopt = compat_sock_common_getsockopt,
1183 #endif
1186 static struct inet_protosw dccp_v6_protosw = {
1187 .type = SOCK_DCCP,
1188 .protocol = IPPROTO_DCCP,
1189 .prot = &dccp_v6_prot,
1190 .ops = &inet6_dccp_ops,
1191 .capability = -1,
1192 .flags = INET_PROTOSW_ICSK,
1195 static int dccp_v6_init_net(struct net *net)
1197 int err;
1199 err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1200 SOCK_DCCP, IPPROTO_DCCP, net);
1201 return err;
1204 static void dccp_v6_exit_net(struct net *net)
1206 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1209 static struct pernet_operations dccp_v6_ops = {
1210 .init = dccp_v6_init_net,
1211 .exit = dccp_v6_exit_net,
1214 static int __init dccp_v6_init(void)
1216 int err = proto_register(&dccp_v6_prot, 1);
1218 if (err != 0)
1219 goto out;
1221 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1222 if (err != 0)
1223 goto out_unregister_proto;
1225 inet6_register_protosw(&dccp_v6_protosw);
1227 err = register_pernet_subsys(&dccp_v6_ops);
1228 if (err != 0)
1229 goto out_destroy_ctl_sock;
1230 out:
1231 return err;
1233 out_destroy_ctl_sock:
1234 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1235 inet6_unregister_protosw(&dccp_v6_protosw);
1236 out_unregister_proto:
1237 proto_unregister(&dccp_v6_prot);
1238 goto out;
1241 static void __exit dccp_v6_exit(void)
1243 unregister_pernet_subsys(&dccp_v6_ops);
1244 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1245 inet6_unregister_protosw(&dccp_v6_protosw);
1246 proto_unregister(&dccp_v6_prot);
1249 module_init(dccp_v6_init);
1250 module_exit(dccp_v6_exit);
1253 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1254 * values directly, Also cover the case where the protocol is not specified,
1255 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1257 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1258 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1259 MODULE_LICENSE("GPL");
1260 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1261 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");