mm-only debug patch...
[mmotm.git] / net / dccp / ipv6.c
bloba2afb553d8b3b9f2ff122f6d50453e7fb24cd5bf
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_v4mapped(newinet->daddr, &newnp->daddr);
515 ipv6_addr_set_v4mapped(newinet->saddr, &newnp->saddr);
517 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
519 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
520 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
521 newnp->pktoptions = NULL;
522 newnp->opt = NULL;
523 newnp->mcast_oif = inet6_iif(skb);
524 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
527 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
528 * here, dccp_create_openreq_child now does this for us, see the comment in
529 * that function for the gory details. -acme
532 /* It is tricky place. Until this moment IPv4 tcp
533 worked with IPv6 icsk.icsk_af_ops.
534 Sync it now.
536 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
538 return newsk;
541 opt = np->opt;
543 if (sk_acceptq_is_full(sk))
544 goto out_overflow;
546 if (dst == NULL) {
547 struct in6_addr *final_p = NULL, final;
548 struct flowi fl;
550 memset(&fl, 0, sizeof(fl));
551 fl.proto = IPPROTO_DCCP;
552 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
553 if (opt != NULL && opt->srcrt != NULL) {
554 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
556 ipv6_addr_copy(&final, &fl.fl6_dst);
557 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
558 final_p = &final;
560 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
561 fl.oif = sk->sk_bound_dev_if;
562 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
563 fl.fl_ip_sport = inet_rsk(req)->loc_port;
564 security_sk_classify_flow(sk, &fl);
566 if (ip6_dst_lookup(sk, &dst, &fl))
567 goto out;
569 if (final_p)
570 ipv6_addr_copy(&fl.fl6_dst, final_p);
572 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
573 goto out;
576 newsk = dccp_create_openreq_child(sk, req, skb);
577 if (newsk == NULL)
578 goto out;
581 * No need to charge this sock to the relevant IPv6 refcnt debug socks
582 * count here, dccp_create_openreq_child now does this for us, see the
583 * comment in that function for the gory details. -acme
586 __ip6_dst_store(newsk, dst, NULL, NULL);
587 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
588 NETIF_F_TSO);
589 newdp6 = (struct dccp6_sock *)newsk;
590 newinet = inet_sk(newsk);
591 newinet->pinet6 = &newdp6->inet6;
592 newdp = dccp_sk(newsk);
593 newnp = inet6_sk(newsk);
595 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
597 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
598 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
599 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
600 newsk->sk_bound_dev_if = ireq6->iif;
602 /* Now IPv6 options...
604 First: no IPv4 options.
606 newinet->opt = NULL;
608 /* Clone RX bits */
609 newnp->rxopt.all = np->rxopt.all;
611 /* Clone pktoptions received with SYN */
612 newnp->pktoptions = NULL;
613 if (ireq6->pktopts != NULL) {
614 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
615 kfree_skb(ireq6->pktopts);
616 ireq6->pktopts = NULL;
617 if (newnp->pktoptions)
618 skb_set_owner_r(newnp->pktoptions, newsk);
620 newnp->opt = NULL;
621 newnp->mcast_oif = inet6_iif(skb);
622 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
625 * Clone native IPv6 options from listening socket (if any)
627 * Yes, keeping reference count would be much more clever, but we make
628 * one more one thing there: reattach optmem to newsk.
630 if (opt != NULL) {
631 newnp->opt = ipv6_dup_options(newsk, opt);
632 if (opt != np->opt)
633 sock_kfree_s(sk, opt, opt->tot_len);
636 inet_csk(newsk)->icsk_ext_hdr_len = 0;
637 if (newnp->opt != NULL)
638 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
639 newnp->opt->opt_flen);
641 dccp_sync_mss(newsk, dst_mtu(dst));
643 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
645 __inet6_hash(newsk);
646 __inet_inherit_port(sk, newsk);
648 return newsk;
650 out_overflow:
651 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
652 out:
653 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
654 if (opt != NULL && opt != np->opt)
655 sock_kfree_s(sk, opt, opt->tot_len);
656 dst_release(dst);
657 return NULL;
660 /* The socket must have it's spinlock held when we get
661 * here.
663 * We have a potential double-lock case here, so even when
664 * doing backlog processing we use the BH locking scheme.
665 * This is because we cannot sleep with the original spinlock
666 * held.
668 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
670 struct ipv6_pinfo *np = inet6_sk(sk);
671 struct sk_buff *opt_skb = NULL;
673 /* Imagine: socket is IPv6. IPv4 packet arrives,
674 goes to IPv4 receive handler and backlogged.
675 From backlog it always goes here. Kerboom...
676 Fortunately, dccp_rcv_established and rcv_established
677 handle them correctly, but it is not case with
678 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
681 if (skb->protocol == htons(ETH_P_IP))
682 return dccp_v4_do_rcv(sk, skb);
684 if (sk_filter(sk, skb))
685 goto discard;
688 * socket locking is here for SMP purposes as backlog rcv is currently
689 * called with bh processing disabled.
692 /* Do Stevens' IPV6_PKTOPTIONS.
694 Yes, guys, it is the only place in our code, where we
695 may make it not affecting IPv4.
696 The rest of code is protocol independent,
697 and I do not like idea to uglify IPv4.
699 Actually, all the idea behind IPV6_PKTOPTIONS
700 looks not very well thought. For now we latch
701 options, received in the last packet, enqueued
702 by tcp. Feel free to propose better solution.
703 --ANK (980728)
705 if (np->rxopt.all)
707 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
708 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
710 opt_skb = skb_clone(skb, GFP_ATOMIC);
712 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
713 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
714 goto reset;
715 if (opt_skb) {
716 /* XXX This is where we would goto ipv6_pktoptions. */
717 __kfree_skb(opt_skb);
719 return 0;
723 * Step 3: Process LISTEN state
724 * If S.state == LISTEN,
725 * If P.type == Request or P contains a valid Init Cookie option,
726 * (* Must scan the packet's options to check for Init
727 * Cookies. Only Init Cookies are processed here,
728 * however; other options are processed in Step 8. This
729 * scan need only be performed if the endpoint uses Init
730 * Cookies *)
731 * (* Generate a new socket and switch to that socket *)
732 * Set S := new socket for this port pair
733 * S.state = RESPOND
734 * Choose S.ISS (initial seqno) or set from Init Cookies
735 * Initialize S.GAR := S.ISS
736 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
737 * Continue with S.state == RESPOND
738 * (* A Response packet will be generated in Step 11 *)
739 * Otherwise,
740 * Generate Reset(No Connection) unless P.type == Reset
741 * Drop packet and return
743 * NOTE: the check for the packet types is done in
744 * dccp_rcv_state_process
746 if (sk->sk_state == DCCP_LISTEN) {
747 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
749 if (nsk == NULL)
750 goto discard;
752 * Queue it on the new socket if the new socket is active,
753 * otherwise we just shortcircuit this and continue with
754 * the new socket..
756 if (nsk != sk) {
757 if (dccp_child_process(sk, nsk, skb))
758 goto reset;
759 if (opt_skb != NULL)
760 __kfree_skb(opt_skb);
761 return 0;
765 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
766 goto reset;
767 if (opt_skb) {
768 /* XXX This is where we would goto ipv6_pktoptions. */
769 __kfree_skb(opt_skb);
771 return 0;
773 reset:
774 dccp_v6_ctl_send_reset(sk, skb);
775 discard:
776 if (opt_skb != NULL)
777 __kfree_skb(opt_skb);
778 kfree_skb(skb);
779 return 0;
782 static int dccp_v6_rcv(struct sk_buff *skb)
784 const struct dccp_hdr *dh;
785 struct sock *sk;
786 int min_cov;
788 /* Step 1: Check header basics */
790 if (dccp_invalid_packet(skb))
791 goto discard_it;
793 /* Step 1: If header checksum is incorrect, drop packet and return. */
794 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
795 &ipv6_hdr(skb)->daddr)) {
796 DCCP_WARN("dropped packet with invalid checksum\n");
797 goto discard_it;
800 dh = dccp_hdr(skb);
802 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
803 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
805 if (dccp_packet_without_ack(skb))
806 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
807 else
808 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
810 /* Step 2:
811 * Look up flow ID in table and get corresponding socket */
812 sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
813 dh->dccph_sport, dh->dccph_dport);
815 * Step 2:
816 * If no socket ...
818 if (sk == NULL) {
819 dccp_pr_debug("failed to look up flow ID in table and "
820 "get corresponding socket\n");
821 goto no_dccp_socket;
825 * Step 2:
826 * ... or S.state == TIMEWAIT,
827 * Generate Reset(No Connection) unless P.type == Reset
828 * Drop packet and return
830 if (sk->sk_state == DCCP_TIME_WAIT) {
831 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
832 inet_twsk_put(inet_twsk(sk));
833 goto no_dccp_socket;
837 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
838 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
839 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
841 min_cov = dccp_sk(sk)->dccps_pcrlen;
842 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
843 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
844 dh->dccph_cscov, min_cov);
845 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
846 goto discard_and_relse;
849 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
850 goto discard_and_relse;
852 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
854 no_dccp_socket:
855 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
856 goto discard_it;
858 * Step 2:
859 * If no socket ...
860 * Generate Reset(No Connection) unless P.type == Reset
861 * Drop packet and return
863 if (dh->dccph_type != DCCP_PKT_RESET) {
864 DCCP_SKB_CB(skb)->dccpd_reset_code =
865 DCCP_RESET_CODE_NO_CONNECTION;
866 dccp_v6_ctl_send_reset(sk, skb);
869 discard_it:
870 kfree_skb(skb);
871 return 0;
873 discard_and_relse:
874 sock_put(sk);
875 goto discard_it;
878 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
879 int addr_len)
881 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
882 struct inet_connection_sock *icsk = inet_csk(sk);
883 struct inet_sock *inet = inet_sk(sk);
884 struct ipv6_pinfo *np = inet6_sk(sk);
885 struct dccp_sock *dp = dccp_sk(sk);
886 struct in6_addr *saddr = NULL, *final_p = NULL, final;
887 struct flowi fl;
888 struct dst_entry *dst;
889 int addr_type;
890 int err;
892 dp->dccps_role = DCCP_ROLE_CLIENT;
894 if (addr_len < SIN6_LEN_RFC2133)
895 return -EINVAL;
897 if (usin->sin6_family != AF_INET6)
898 return -EAFNOSUPPORT;
900 memset(&fl, 0, sizeof(fl));
902 if (np->sndflow) {
903 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
904 IP6_ECN_flow_init(fl.fl6_flowlabel);
905 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
906 struct ip6_flowlabel *flowlabel;
907 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
908 if (flowlabel == NULL)
909 return -EINVAL;
910 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
911 fl6_sock_release(flowlabel);
915 * connect() to INADDR_ANY means loopback (BSD'ism).
917 if (ipv6_addr_any(&usin->sin6_addr))
918 usin->sin6_addr.s6_addr[15] = 1;
920 addr_type = ipv6_addr_type(&usin->sin6_addr);
922 if (addr_type & IPV6_ADDR_MULTICAST)
923 return -ENETUNREACH;
925 if (addr_type & IPV6_ADDR_LINKLOCAL) {
926 if (addr_len >= sizeof(struct sockaddr_in6) &&
927 usin->sin6_scope_id) {
928 /* If interface is set while binding, indices
929 * must coincide.
931 if (sk->sk_bound_dev_if &&
932 sk->sk_bound_dev_if != usin->sin6_scope_id)
933 return -EINVAL;
935 sk->sk_bound_dev_if = usin->sin6_scope_id;
938 /* Connect to link-local address requires an interface */
939 if (!sk->sk_bound_dev_if)
940 return -EINVAL;
943 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
944 np->flow_label = fl.fl6_flowlabel;
947 * DCCP over IPv4
949 if (addr_type == IPV6_ADDR_MAPPED) {
950 u32 exthdrlen = icsk->icsk_ext_hdr_len;
951 struct sockaddr_in sin;
953 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
955 if (__ipv6_only_sock(sk))
956 return -ENETUNREACH;
958 sin.sin_family = AF_INET;
959 sin.sin_port = usin->sin6_port;
960 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
962 icsk->icsk_af_ops = &dccp_ipv6_mapped;
963 sk->sk_backlog_rcv = dccp_v4_do_rcv;
965 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
966 if (err) {
967 icsk->icsk_ext_hdr_len = exthdrlen;
968 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
969 sk->sk_backlog_rcv = dccp_v6_do_rcv;
970 goto failure;
971 } else {
972 ipv6_addr_set_v4mapped(inet->saddr, &np->saddr);
973 ipv6_addr_set_v4mapped(inet->rcv_saddr, &np->rcv_saddr);
976 return err;
979 if (!ipv6_addr_any(&np->rcv_saddr))
980 saddr = &np->rcv_saddr;
982 fl.proto = IPPROTO_DCCP;
983 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
984 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
985 fl.oif = sk->sk_bound_dev_if;
986 fl.fl_ip_dport = usin->sin6_port;
987 fl.fl_ip_sport = inet->sport;
988 security_sk_classify_flow(sk, &fl);
990 if (np->opt != NULL && np->opt->srcrt != NULL) {
991 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
993 ipv6_addr_copy(&final, &fl.fl6_dst);
994 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
995 final_p = &final;
998 err = ip6_dst_lookup(sk, &dst, &fl);
999 if (err)
1000 goto failure;
1002 if (final_p)
1003 ipv6_addr_copy(&fl.fl6_dst, final_p);
1005 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
1006 if (err < 0) {
1007 if (err == -EREMOTE)
1008 err = ip6_dst_blackhole(sk, &dst, &fl);
1009 if (err < 0)
1010 goto failure;
1013 if (saddr == NULL) {
1014 saddr = &fl.fl6_src;
1015 ipv6_addr_copy(&np->rcv_saddr, saddr);
1018 /* set the source address */
1019 ipv6_addr_copy(&np->saddr, saddr);
1020 inet->rcv_saddr = LOOPBACK4_IPV6;
1022 __ip6_dst_store(sk, dst, NULL, NULL);
1024 icsk->icsk_ext_hdr_len = 0;
1025 if (np->opt != NULL)
1026 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1027 np->opt->opt_nflen);
1029 inet->dport = usin->sin6_port;
1031 dccp_set_state(sk, DCCP_REQUESTING);
1032 err = inet6_hash_connect(&dccp_death_row, sk);
1033 if (err)
1034 goto late_failure;
1036 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1037 np->daddr.s6_addr32,
1038 inet->sport, inet->dport);
1039 err = dccp_connect(sk);
1040 if (err)
1041 goto late_failure;
1043 return 0;
1045 late_failure:
1046 dccp_set_state(sk, DCCP_CLOSED);
1047 __sk_dst_reset(sk);
1048 failure:
1049 inet->dport = 0;
1050 sk->sk_route_caps = 0;
1051 return err;
1054 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1055 .queue_xmit = inet6_csk_xmit,
1056 .send_check = dccp_v6_send_check,
1057 .rebuild_header = inet6_sk_rebuild_header,
1058 .conn_request = dccp_v6_conn_request,
1059 .syn_recv_sock = dccp_v6_request_recv_sock,
1060 .net_header_len = sizeof(struct ipv6hdr),
1061 .setsockopt = ipv6_setsockopt,
1062 .getsockopt = ipv6_getsockopt,
1063 .addr2sockaddr = inet6_csk_addr2sockaddr,
1064 .sockaddr_len = sizeof(struct sockaddr_in6),
1065 .bind_conflict = inet6_csk_bind_conflict,
1066 #ifdef CONFIG_COMPAT
1067 .compat_setsockopt = compat_ipv6_setsockopt,
1068 .compat_getsockopt = compat_ipv6_getsockopt,
1069 #endif
1073 * DCCP over IPv4 via INET6 API
1075 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1076 .queue_xmit = ip_queue_xmit,
1077 .send_check = dccp_v4_send_check,
1078 .rebuild_header = inet_sk_rebuild_header,
1079 .conn_request = dccp_v6_conn_request,
1080 .syn_recv_sock = dccp_v6_request_recv_sock,
1081 .net_header_len = sizeof(struct iphdr),
1082 .setsockopt = ipv6_setsockopt,
1083 .getsockopt = ipv6_getsockopt,
1084 .addr2sockaddr = inet6_csk_addr2sockaddr,
1085 .sockaddr_len = sizeof(struct sockaddr_in6),
1086 #ifdef CONFIG_COMPAT
1087 .compat_setsockopt = compat_ipv6_setsockopt,
1088 .compat_getsockopt = compat_ipv6_getsockopt,
1089 #endif
1092 /* NOTE: A lot of things set to zero explicitly by call to
1093 * sk_alloc() so need not be done here.
1095 static int dccp_v6_init_sock(struct sock *sk)
1097 static __u8 dccp_v6_ctl_sock_initialized;
1098 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1100 if (err == 0) {
1101 if (unlikely(!dccp_v6_ctl_sock_initialized))
1102 dccp_v6_ctl_sock_initialized = 1;
1103 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1106 return err;
1109 static void dccp_v6_destroy_sock(struct sock *sk)
1111 dccp_destroy_sock(sk);
1112 inet6_destroy_sock(sk);
1115 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1116 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1119 static struct proto dccp_v6_prot = {
1120 .name = "DCCPv6",
1121 .owner = THIS_MODULE,
1122 .close = dccp_close,
1123 .connect = dccp_v6_connect,
1124 .disconnect = dccp_disconnect,
1125 .ioctl = dccp_ioctl,
1126 .init = dccp_v6_init_sock,
1127 .setsockopt = dccp_setsockopt,
1128 .getsockopt = dccp_getsockopt,
1129 .sendmsg = dccp_sendmsg,
1130 .recvmsg = dccp_recvmsg,
1131 .backlog_rcv = dccp_v6_do_rcv,
1132 .hash = dccp_v6_hash,
1133 .unhash = inet_unhash,
1134 .accept = inet_csk_accept,
1135 .get_port = inet_csk_get_port,
1136 .shutdown = dccp_shutdown,
1137 .destroy = dccp_v6_destroy_sock,
1138 .orphan_count = &dccp_orphan_count,
1139 .max_header = MAX_DCCP_HEADER,
1140 .obj_size = sizeof(struct dccp6_sock),
1141 .slab_flags = SLAB_DESTROY_BY_RCU,
1142 .rsk_prot = &dccp6_request_sock_ops,
1143 .twsk_prot = &dccp6_timewait_sock_ops,
1144 .h.hashinfo = &dccp_hashinfo,
1145 #ifdef CONFIG_COMPAT
1146 .compat_setsockopt = compat_dccp_setsockopt,
1147 .compat_getsockopt = compat_dccp_getsockopt,
1148 #endif
1151 static const struct inet6_protocol dccp_v6_protocol = {
1152 .handler = dccp_v6_rcv,
1153 .err_handler = dccp_v6_err,
1154 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1157 static const struct proto_ops inet6_dccp_ops = {
1158 .family = PF_INET6,
1159 .owner = THIS_MODULE,
1160 .release = inet6_release,
1161 .bind = inet6_bind,
1162 .connect = inet_stream_connect,
1163 .socketpair = sock_no_socketpair,
1164 .accept = inet_accept,
1165 .getname = inet6_getname,
1166 .poll = dccp_poll,
1167 .ioctl = inet6_ioctl,
1168 .listen = inet_dccp_listen,
1169 .shutdown = inet_shutdown,
1170 .setsockopt = sock_common_setsockopt,
1171 .getsockopt = sock_common_getsockopt,
1172 .sendmsg = inet_sendmsg,
1173 .recvmsg = sock_common_recvmsg,
1174 .mmap = sock_no_mmap,
1175 .sendpage = sock_no_sendpage,
1176 #ifdef CONFIG_COMPAT
1177 .compat_setsockopt = compat_sock_common_setsockopt,
1178 .compat_getsockopt = compat_sock_common_getsockopt,
1179 #endif
1182 static struct inet_protosw dccp_v6_protosw = {
1183 .type = SOCK_DCCP,
1184 .protocol = IPPROTO_DCCP,
1185 .prot = &dccp_v6_prot,
1186 .ops = &inet6_dccp_ops,
1187 .capability = -1,
1188 .flags = INET_PROTOSW_ICSK,
1191 static int dccp_v6_init_net(struct net *net)
1193 int err;
1195 err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1196 SOCK_DCCP, IPPROTO_DCCP, net);
1197 return err;
1200 static void dccp_v6_exit_net(struct net *net)
1202 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1205 static struct pernet_operations dccp_v6_ops = {
1206 .init = dccp_v6_init_net,
1207 .exit = dccp_v6_exit_net,
1210 static int __init dccp_v6_init(void)
1212 int err = proto_register(&dccp_v6_prot, 1);
1214 if (err != 0)
1215 goto out;
1217 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1218 if (err != 0)
1219 goto out_unregister_proto;
1221 inet6_register_protosw(&dccp_v6_protosw);
1223 err = register_pernet_subsys(&dccp_v6_ops);
1224 if (err != 0)
1225 goto out_destroy_ctl_sock;
1226 out:
1227 return err;
1229 out_destroy_ctl_sock:
1230 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1231 inet6_unregister_protosw(&dccp_v6_protosw);
1232 out_unregister_proto:
1233 proto_unregister(&dccp_v6_prot);
1234 goto out;
1237 static void __exit dccp_v6_exit(void)
1239 unregister_pernet_subsys(&dccp_v6_ops);
1240 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1241 inet6_unregister_protosw(&dccp_v6_protosw);
1242 proto_unregister(&dccp_v6_prot);
1245 module_init(dccp_v6_init);
1246 module_exit(dccp_v6_exit);
1249 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1250 * values directly, Also cover the case where the protocol is not specified,
1251 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1253 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1254 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1255 MODULE_LICENSE("GPL");
1256 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1257 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");