writeback: split writeback_inodes_wb
[linux-2.6/next.git] / net / dccp / ipv6.c
blob091698899594388bbfb65678ce2ca8d1061283aa
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/slab.h>
18 #include <linux/xfrm.h>
20 #include <net/addrconf.h>
21 #include <net/inet_common.h>
22 #include <net/inet_hashtables.h>
23 #include <net/inet_sock.h>
24 #include <net/inet6_connection_sock.h>
25 #include <net/inet6_hashtables.h>
26 #include <net/ip6_route.h>
27 #include <net/ipv6.h>
28 #include <net/protocol.h>
29 #include <net/transp_v6.h>
30 #include <net/ip6_checksum.h>
31 #include <net/xfrm.h>
33 #include "dccp.h"
34 #include "ipv6.h"
35 #include "feat.h"
37 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
39 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
40 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
42 static void dccp_v6_hash(struct sock *sk)
44 if (sk->sk_state != DCCP_CLOSED) {
45 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
46 inet_hash(sk);
47 return;
49 local_bh_disable();
50 __inet6_hash(sk, NULL);
51 local_bh_enable();
55 /* add pseudo-header to DCCP checksum stored in skb->csum */
56 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
57 struct in6_addr *saddr,
58 struct in6_addr *daddr)
60 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
63 static inline void dccp_v6_send_check(struct sock *sk, 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->inet_dport;
162 fl.fl_ip_sport = inet->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,
245 struct request_values *rv_unused)
247 struct inet6_request_sock *ireq6 = inet6_rsk(req);
248 struct ipv6_pinfo *np = inet6_sk(sk);
249 struct sk_buff *skb;
250 struct ipv6_txoptions *opt = NULL;
251 struct in6_addr *final_p = NULL, final;
252 struct flowi fl;
253 int err = -1;
254 struct dst_entry *dst;
256 memset(&fl, 0, sizeof(fl));
257 fl.proto = IPPROTO_DCCP;
258 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
259 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
260 fl.fl6_flowlabel = 0;
261 fl.oif = ireq6->iif;
262 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
263 fl.fl_ip_sport = inet_rsk(req)->loc_port;
264 security_req_classify_flow(req, &fl);
266 opt = np->opt;
268 if (opt != NULL && opt->srcrt != NULL) {
269 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
271 ipv6_addr_copy(&final, &fl.fl6_dst);
272 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
273 final_p = &final;
276 err = ip6_dst_lookup(sk, &dst, &fl);
277 if (err)
278 goto done;
280 if (final_p)
281 ipv6_addr_copy(&fl.fl6_dst, final_p);
283 err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0);
284 if (err < 0)
285 goto done;
287 skb = dccp_make_response(sk, dst, req);
288 if (skb != NULL) {
289 struct dccp_hdr *dh = dccp_hdr(skb);
291 dh->dccph_checksum = dccp_v6_csum_finish(skb,
292 &ireq6->loc_addr,
293 &ireq6->rmt_addr);
294 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
295 err = ip6_xmit(sk, skb, &fl, opt);
296 err = net_xmit_eval(err);
299 done:
300 if (opt != NULL && opt != np->opt)
301 sock_kfree_s(sk, opt, opt->tot_len);
302 dst_release(dst);
303 return err;
306 static void dccp_v6_reqsk_destructor(struct request_sock *req)
308 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
309 if (inet6_rsk(req)->pktopts != NULL)
310 kfree_skb(inet6_rsk(req)->pktopts);
313 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
315 struct ipv6hdr *rxip6h;
316 struct sk_buff *skb;
317 struct flowi fl;
318 struct net *net = dev_net(skb_dst(rxskb)->dev);
319 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
320 struct dst_entry *dst;
322 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
323 return;
325 if (!ipv6_unicast_destination(rxskb))
326 return;
328 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
329 if (skb == NULL)
330 return;
332 rxip6h = ipv6_hdr(rxskb);
333 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
334 &rxip6h->daddr);
336 memset(&fl, 0, sizeof(fl));
337 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
338 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
340 fl.proto = IPPROTO_DCCP;
341 fl.oif = inet6_iif(rxskb);
342 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
343 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
344 security_skb_classify_flow(rxskb, &fl);
346 /* sk = NULL, but it is safe for now. RST socket required. */
347 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) {
348 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) {
349 skb_dst_set(skb, dst);
350 ip6_xmit(ctl_sk, skb, &fl, NULL);
351 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
352 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
353 return;
357 kfree_skb(skb);
360 static struct request_sock_ops dccp6_request_sock_ops = {
361 .family = AF_INET6,
362 .obj_size = sizeof(struct dccp6_request_sock),
363 .rtx_syn_ack = dccp_v6_send_response,
364 .send_ack = dccp_reqsk_send_ack,
365 .destructor = dccp_v6_reqsk_destructor,
366 .send_reset = dccp_v6_ctl_send_reset,
369 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
371 const struct dccp_hdr *dh = dccp_hdr(skb);
372 const struct ipv6hdr *iph = ipv6_hdr(skb);
373 struct sock *nsk;
374 struct request_sock **prev;
375 /* Find possible connection requests. */
376 struct request_sock *req = inet6_csk_search_req(sk, &prev,
377 dh->dccph_sport,
378 &iph->saddr,
379 &iph->daddr,
380 inet6_iif(skb));
381 if (req != NULL)
382 return dccp_check_req(sk, skb, req, prev);
384 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
385 &iph->saddr, dh->dccph_sport,
386 &iph->daddr, ntohs(dh->dccph_dport),
387 inet6_iif(skb));
388 if (nsk != NULL) {
389 if (nsk->sk_state != DCCP_TIME_WAIT) {
390 bh_lock_sock(nsk);
391 return nsk;
393 inet_twsk_put(inet_twsk(nsk));
394 return NULL;
397 return sk;
400 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
402 struct request_sock *req;
403 struct dccp_request_sock *dreq;
404 struct inet6_request_sock *ireq6;
405 struct ipv6_pinfo *np = inet6_sk(sk);
406 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
407 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
409 if (skb->protocol == htons(ETH_P_IP))
410 return dccp_v4_conn_request(sk, skb);
412 if (!ipv6_unicast_destination(skb))
413 return 0; /* discard, don't send a reset here */
415 if (dccp_bad_service_code(sk, service)) {
416 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
417 goto drop;
420 * There are no SYN attacks on IPv6, yet...
422 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
423 if (inet_csk_reqsk_queue_is_full(sk))
424 goto drop;
426 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
427 goto drop;
429 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
430 if (req == NULL)
431 goto drop;
433 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
434 goto drop_and_free;
436 dreq = dccp_rsk(req);
437 if (dccp_parse_options(sk, dreq, skb))
438 goto drop_and_free;
440 if (security_inet_conn_request(sk, skb, req))
441 goto drop_and_free;
443 ireq6 = inet6_rsk(req);
444 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
445 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
447 if (ipv6_opt_accepted(sk, skb) ||
448 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
449 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
450 atomic_inc(&skb->users);
451 ireq6->pktopts = skb;
453 ireq6->iif = sk->sk_bound_dev_if;
455 /* So that link locals have meaning */
456 if (!sk->sk_bound_dev_if &&
457 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
458 ireq6->iif = inet6_iif(skb);
461 * Step 3: Process LISTEN state
463 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
465 * In fact we defer setting S.GSR, S.SWL, S.SWH to
466 * dccp_create_openreq_child.
468 dreq->dreq_isr = dcb->dccpd_seq;
469 dreq->dreq_iss = dccp_v6_init_sequence(skb);
470 dreq->dreq_service = service;
472 if (dccp_v6_send_response(sk, req, NULL))
473 goto drop_and_free;
475 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
476 return 0;
478 drop_and_free:
479 reqsk_free(req);
480 drop:
481 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
482 return -1;
485 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
486 struct sk_buff *skb,
487 struct request_sock *req,
488 struct dst_entry *dst)
490 struct inet6_request_sock *ireq6 = inet6_rsk(req);
491 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
492 struct inet_sock *newinet;
493 struct dccp_sock *newdp;
494 struct dccp6_sock *newdp6;
495 struct sock *newsk;
496 struct ipv6_txoptions *opt;
498 if (skb->protocol == htons(ETH_P_IP)) {
500 * v6 mapped
502 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
503 if (newsk == NULL)
504 return NULL;
506 newdp6 = (struct dccp6_sock *)newsk;
507 newdp = dccp_sk(newsk);
508 newinet = inet_sk(newsk);
509 newinet->pinet6 = &newdp6->inet6;
510 newnp = inet6_sk(newsk);
512 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
514 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
516 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
518 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
520 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
521 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
522 newnp->pktoptions = NULL;
523 newnp->opt = NULL;
524 newnp->mcast_oif = inet6_iif(skb);
525 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
528 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
529 * here, dccp_create_openreq_child now does this for us, see the comment in
530 * that function for the gory details. -acme
533 /* It is tricky place. Until this moment IPv4 tcp
534 worked with IPv6 icsk.icsk_af_ops.
535 Sync it now.
537 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
539 return newsk;
542 opt = np->opt;
544 if (sk_acceptq_is_full(sk))
545 goto out_overflow;
547 if (dst == NULL) {
548 struct in6_addr *final_p = NULL, final;
549 struct flowi fl;
551 memset(&fl, 0, sizeof(fl));
552 fl.proto = IPPROTO_DCCP;
553 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
554 if (opt != NULL && opt->srcrt != NULL) {
555 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
557 ipv6_addr_copy(&final, &fl.fl6_dst);
558 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
559 final_p = &final;
561 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
562 fl.oif = sk->sk_bound_dev_if;
563 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
564 fl.fl_ip_sport = inet_rsk(req)->loc_port;
565 security_sk_classify_flow(sk, &fl);
567 if (ip6_dst_lookup(sk, &dst, &fl))
568 goto out;
570 if (final_p)
571 ipv6_addr_copy(&fl.fl6_dst, final_p);
573 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
574 goto out;
577 newsk = dccp_create_openreq_child(sk, req, skb);
578 if (newsk == NULL)
579 goto out;
582 * No need to charge this sock to the relevant IPv6 refcnt debug socks
583 * count here, dccp_create_openreq_child now does this for us, see the
584 * comment in that function for the gory details. -acme
587 __ip6_dst_store(newsk, dst, NULL, NULL);
588 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
589 NETIF_F_TSO);
590 newdp6 = (struct dccp6_sock *)newsk;
591 newinet = inet_sk(newsk);
592 newinet->pinet6 = &newdp6->inet6;
593 newdp = dccp_sk(newsk);
594 newnp = inet6_sk(newsk);
596 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
598 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
599 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
600 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
601 newsk->sk_bound_dev_if = ireq6->iif;
603 /* Now IPv6 options...
605 First: no IPv4 options.
607 newinet->opt = NULL;
609 /* Clone RX bits */
610 newnp->rxopt.all = np->rxopt.all;
612 /* Clone pktoptions received with SYN */
613 newnp->pktoptions = NULL;
614 if (ireq6->pktopts != NULL) {
615 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
616 kfree_skb(ireq6->pktopts);
617 ireq6->pktopts = NULL;
618 if (newnp->pktoptions)
619 skb_set_owner_r(newnp->pktoptions, newsk);
621 newnp->opt = NULL;
622 newnp->mcast_oif = inet6_iif(skb);
623 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
626 * Clone native IPv6 options from listening socket (if any)
628 * Yes, keeping reference count would be much more clever, but we make
629 * one more one thing there: reattach optmem to newsk.
631 if (opt != NULL) {
632 newnp->opt = ipv6_dup_options(newsk, opt);
633 if (opt != np->opt)
634 sock_kfree_s(sk, opt, opt->tot_len);
637 inet_csk(newsk)->icsk_ext_hdr_len = 0;
638 if (newnp->opt != NULL)
639 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
640 newnp->opt->opt_flen);
642 dccp_sync_mss(newsk, dst_mtu(dst));
644 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
645 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
647 __inet6_hash(newsk, NULL);
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;
974 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
975 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr);
977 return err;
980 if (!ipv6_addr_any(&np->rcv_saddr))
981 saddr = &np->rcv_saddr;
983 fl.proto = IPPROTO_DCCP;
984 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
985 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
986 fl.oif = sk->sk_bound_dev_if;
987 fl.fl_ip_dport = usin->sin6_port;
988 fl.fl_ip_sport = inet->inet_sport;
989 security_sk_classify_flow(sk, &fl);
991 if (np->opt != NULL && np->opt->srcrt != NULL) {
992 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
994 ipv6_addr_copy(&final, &fl.fl6_dst);
995 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
996 final_p = &final;
999 err = ip6_dst_lookup(sk, &dst, &fl);
1000 if (err)
1001 goto failure;
1003 if (final_p)
1004 ipv6_addr_copy(&fl.fl6_dst, final_p);
1006 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
1007 if (err < 0) {
1008 if (err == -EREMOTE)
1009 err = ip6_dst_blackhole(sk, &dst, &fl);
1010 if (err < 0)
1011 goto failure;
1014 if (saddr == NULL) {
1015 saddr = &fl.fl6_src;
1016 ipv6_addr_copy(&np->rcv_saddr, saddr);
1019 /* set the source address */
1020 ipv6_addr_copy(&np->saddr, saddr);
1021 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
1023 __ip6_dst_store(sk, dst, NULL, NULL);
1025 icsk->icsk_ext_hdr_len = 0;
1026 if (np->opt != NULL)
1027 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1028 np->opt->opt_nflen);
1030 inet->inet_dport = usin->sin6_port;
1032 dccp_set_state(sk, DCCP_REQUESTING);
1033 err = inet6_hash_connect(&dccp_death_row, sk);
1034 if (err)
1035 goto late_failure;
1037 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1038 np->daddr.s6_addr32,
1039 inet->inet_sport,
1040 inet->inet_dport);
1041 err = dccp_connect(sk);
1042 if (err)
1043 goto late_failure;
1045 return 0;
1047 late_failure:
1048 dccp_set_state(sk, DCCP_CLOSED);
1049 __sk_dst_reset(sk);
1050 failure:
1051 inet->inet_dport = 0;
1052 sk->sk_route_caps = 0;
1053 return err;
1056 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1057 .queue_xmit = inet6_csk_xmit,
1058 .send_check = dccp_v6_send_check,
1059 .rebuild_header = inet6_sk_rebuild_header,
1060 .conn_request = dccp_v6_conn_request,
1061 .syn_recv_sock = dccp_v6_request_recv_sock,
1062 .net_header_len = sizeof(struct ipv6hdr),
1063 .setsockopt = ipv6_setsockopt,
1064 .getsockopt = ipv6_getsockopt,
1065 .addr2sockaddr = inet6_csk_addr2sockaddr,
1066 .sockaddr_len = sizeof(struct sockaddr_in6),
1067 .bind_conflict = inet6_csk_bind_conflict,
1068 #ifdef CONFIG_COMPAT
1069 .compat_setsockopt = compat_ipv6_setsockopt,
1070 .compat_getsockopt = compat_ipv6_getsockopt,
1071 #endif
1075 * DCCP over IPv4 via INET6 API
1077 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1078 .queue_xmit = ip_queue_xmit,
1079 .send_check = dccp_v4_send_check,
1080 .rebuild_header = inet_sk_rebuild_header,
1081 .conn_request = dccp_v6_conn_request,
1082 .syn_recv_sock = dccp_v6_request_recv_sock,
1083 .net_header_len = sizeof(struct iphdr),
1084 .setsockopt = ipv6_setsockopt,
1085 .getsockopt = ipv6_getsockopt,
1086 .addr2sockaddr = inet6_csk_addr2sockaddr,
1087 .sockaddr_len = sizeof(struct sockaddr_in6),
1088 #ifdef CONFIG_COMPAT
1089 .compat_setsockopt = compat_ipv6_setsockopt,
1090 .compat_getsockopt = compat_ipv6_getsockopt,
1091 #endif
1094 /* NOTE: A lot of things set to zero explicitly by call to
1095 * sk_alloc() so need not be done here.
1097 static int dccp_v6_init_sock(struct sock *sk)
1099 static __u8 dccp_v6_ctl_sock_initialized;
1100 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1102 if (err == 0) {
1103 if (unlikely(!dccp_v6_ctl_sock_initialized))
1104 dccp_v6_ctl_sock_initialized = 1;
1105 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1108 return err;
1111 static void dccp_v6_destroy_sock(struct sock *sk)
1113 dccp_destroy_sock(sk);
1114 inet6_destroy_sock(sk);
1117 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1118 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1121 static struct proto dccp_v6_prot = {
1122 .name = "DCCPv6",
1123 .owner = THIS_MODULE,
1124 .close = dccp_close,
1125 .connect = dccp_v6_connect,
1126 .disconnect = dccp_disconnect,
1127 .ioctl = dccp_ioctl,
1128 .init = dccp_v6_init_sock,
1129 .setsockopt = dccp_setsockopt,
1130 .getsockopt = dccp_getsockopt,
1131 .sendmsg = dccp_sendmsg,
1132 .recvmsg = dccp_recvmsg,
1133 .backlog_rcv = dccp_v6_do_rcv,
1134 .hash = dccp_v6_hash,
1135 .unhash = inet_unhash,
1136 .accept = inet_csk_accept,
1137 .get_port = inet_csk_get_port,
1138 .shutdown = dccp_shutdown,
1139 .destroy = dccp_v6_destroy_sock,
1140 .orphan_count = &dccp_orphan_count,
1141 .max_header = MAX_DCCP_HEADER,
1142 .obj_size = sizeof(struct dccp6_sock),
1143 .slab_flags = SLAB_DESTROY_BY_RCU,
1144 .rsk_prot = &dccp6_request_sock_ops,
1145 .twsk_prot = &dccp6_timewait_sock_ops,
1146 .h.hashinfo = &dccp_hashinfo,
1147 #ifdef CONFIG_COMPAT
1148 .compat_setsockopt = compat_dccp_setsockopt,
1149 .compat_getsockopt = compat_dccp_getsockopt,
1150 #endif
1153 static const struct inet6_protocol dccp_v6_protocol = {
1154 .handler = dccp_v6_rcv,
1155 .err_handler = dccp_v6_err,
1156 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1159 static const struct proto_ops inet6_dccp_ops = {
1160 .family = PF_INET6,
1161 .owner = THIS_MODULE,
1162 .release = inet6_release,
1163 .bind = inet6_bind,
1164 .connect = inet_stream_connect,
1165 .socketpair = sock_no_socketpair,
1166 .accept = inet_accept,
1167 .getname = inet6_getname,
1168 .poll = dccp_poll,
1169 .ioctl = inet6_ioctl,
1170 .listen = inet_dccp_listen,
1171 .shutdown = inet_shutdown,
1172 .setsockopt = sock_common_setsockopt,
1173 .getsockopt = sock_common_getsockopt,
1174 .sendmsg = inet_sendmsg,
1175 .recvmsg = sock_common_recvmsg,
1176 .mmap = sock_no_mmap,
1177 .sendpage = sock_no_sendpage,
1178 #ifdef CONFIG_COMPAT
1179 .compat_setsockopt = compat_sock_common_setsockopt,
1180 .compat_getsockopt = compat_sock_common_getsockopt,
1181 #endif
1184 static struct inet_protosw dccp_v6_protosw = {
1185 .type = SOCK_DCCP,
1186 .protocol = IPPROTO_DCCP,
1187 .prot = &dccp_v6_prot,
1188 .ops = &inet6_dccp_ops,
1189 .flags = INET_PROTOSW_ICSK,
1192 static int __net_init dccp_v6_init_net(struct net *net)
1194 if (dccp_hashinfo.bhash == NULL)
1195 return -ESOCKTNOSUPPORT;
1197 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1198 SOCK_DCCP, IPPROTO_DCCP, net);
1201 static void __net_exit dccp_v6_exit_net(struct net *net)
1203 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1206 static struct pernet_operations dccp_v6_ops = {
1207 .init = dccp_v6_init_net,
1208 .exit = dccp_v6_exit_net,
1211 static int __init dccp_v6_init(void)
1213 int err = proto_register(&dccp_v6_prot, 1);
1215 if (err != 0)
1216 goto out;
1218 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1219 if (err != 0)
1220 goto out_unregister_proto;
1222 inet6_register_protosw(&dccp_v6_protosw);
1224 err = register_pernet_subsys(&dccp_v6_ops);
1225 if (err != 0)
1226 goto out_destroy_ctl_sock;
1227 out:
1228 return err;
1230 out_destroy_ctl_sock:
1231 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1232 inet6_unregister_protosw(&dccp_v6_protosw);
1233 out_unregister_proto:
1234 proto_unregister(&dccp_v6_prot);
1235 goto out;
1238 static void __exit dccp_v6_exit(void)
1240 unregister_pernet_subsys(&dccp_v6_ops);
1241 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1242 inet6_unregister_protosw(&dccp_v6_protosw);
1243 proto_unregister(&dccp_v6_prot);
1246 module_init(dccp_v6_init);
1247 module_exit(dccp_v6_exit);
1250 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1251 * values directly, Also cover the case where the protocol is not specified,
1252 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1254 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1255 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1256 MODULE_LICENSE("GPL");
1257 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1258 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");