mm: fix page table unmap for stack guard page properly
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / dccp / ipv6.c
blob3b11e41a2929e717ded1865c864a6f7b1b5fb7e0
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, int unused_value,
64 struct sk_buff *skb)
66 struct ipv6_pinfo *np = inet6_sk(sk);
67 struct dccp_hdr *dh = dccp_hdr(skb);
69 dccp_csum_outgoing(skb);
70 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr);
73 static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
74 __be16 sport, __be16 dport )
76 return secure_tcpv6_sequence_number(saddr, daddr, sport, dport);
79 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb)
81 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
82 ipv6_hdr(skb)->saddr.s6_addr32,
83 dccp_hdr(skb)->dccph_dport,
84 dccp_hdr(skb)->dccph_sport );
88 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
89 u8 type, u8 code, int offset, __be32 info)
91 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data;
92 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
93 struct dccp_sock *dp;
94 struct ipv6_pinfo *np;
95 struct sock *sk;
96 int err;
97 __u64 seq;
98 struct net *net = dev_net(skb->dev);
100 if (skb->len < offset + sizeof(*dh) ||
101 skb->len < offset + __dccp_basic_hdr_len(dh)) {
102 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
103 ICMP6_MIB_INERRORS);
104 return;
107 sk = inet6_lookup(net, &dccp_hashinfo,
108 &hdr->daddr, dh->dccph_dport,
109 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
111 if (sk == NULL) {
112 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
113 ICMP6_MIB_INERRORS);
114 return;
117 if (sk->sk_state == DCCP_TIME_WAIT) {
118 inet_twsk_put(inet_twsk(sk));
119 return;
122 bh_lock_sock(sk);
123 if (sock_owned_by_user(sk))
124 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
126 if (sk->sk_state == DCCP_CLOSED)
127 goto out;
129 dp = dccp_sk(sk);
130 seq = dccp_hdr_seq(dh);
131 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
132 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
133 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
134 goto out;
137 np = inet6_sk(sk);
139 if (type == ICMPV6_PKT_TOOBIG) {
140 struct dst_entry *dst = NULL;
142 if (sock_owned_by_user(sk))
143 goto out;
144 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
145 goto out;
147 /* icmp should have updated the destination cache entry */
148 dst = __sk_dst_check(sk, np->dst_cookie);
149 if (dst == NULL) {
150 struct inet_sock *inet = inet_sk(sk);
151 struct flowi fl;
153 /* BUGGG_FUTURE: Again, it is not clear how
154 to handle rthdr case. Ignore this complexity
155 for now.
157 memset(&fl, 0, sizeof(fl));
158 fl.proto = IPPROTO_DCCP;
159 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
160 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
161 fl.oif = sk->sk_bound_dev_if;
162 fl.fl_ip_dport = inet->inet_dport;
163 fl.fl_ip_sport = inet->inet_sport;
164 security_sk_classify_flow(sk, &fl);
166 err = ip6_dst_lookup(sk, &dst, &fl);
167 if (err) {
168 sk->sk_err_soft = -err;
169 goto out;
172 err = xfrm_lookup(net, &dst, &fl, sk, 0);
173 if (err < 0) {
174 sk->sk_err_soft = -err;
175 goto out;
177 } else
178 dst_hold(dst);
180 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
181 dccp_sync_mss(sk, dst_mtu(dst));
182 } /* else let the usual retransmit timer handle it */
183 dst_release(dst);
184 goto out;
187 icmpv6_err_convert(type, code, &err);
189 /* Might be for an request_sock */
190 switch (sk->sk_state) {
191 struct request_sock *req, **prev;
192 case DCCP_LISTEN:
193 if (sock_owned_by_user(sk))
194 goto out;
196 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
197 &hdr->daddr, &hdr->saddr,
198 inet6_iif(skb));
199 if (req == NULL)
200 goto out;
203 * ICMPs are not backlogged, hence we cannot get an established
204 * socket here.
206 WARN_ON(req->sk != NULL);
208 if (seq != dccp_rsk(req)->dreq_iss) {
209 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
210 goto out;
213 inet_csk_reqsk_queue_drop(sk, req, prev);
214 goto out;
216 case DCCP_REQUESTING:
217 case DCCP_RESPOND: /* Cannot happen.
218 It can, it SYNs are crossed. --ANK */
219 if (!sock_owned_by_user(sk)) {
220 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
221 sk->sk_err = err;
223 * Wake people up to see the error
224 * (see connect in sock.c)
226 sk->sk_error_report(sk);
227 dccp_done(sk);
228 } else
229 sk->sk_err_soft = err;
230 goto out;
233 if (!sock_owned_by_user(sk) && np->recverr) {
234 sk->sk_err = err;
235 sk->sk_error_report(sk);
236 } else
237 sk->sk_err_soft = err;
239 out:
240 bh_unlock_sock(sk);
241 sock_put(sk);
245 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
246 struct request_values *rv_unused)
248 struct inet6_request_sock *ireq6 = inet6_rsk(req);
249 struct ipv6_pinfo *np = inet6_sk(sk);
250 struct sk_buff *skb;
251 struct ipv6_txoptions *opt = NULL;
252 struct in6_addr *final_p = NULL, final;
253 struct flowi fl;
254 int err = -1;
255 struct dst_entry *dst;
257 memset(&fl, 0, sizeof(fl));
258 fl.proto = IPPROTO_DCCP;
259 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
260 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
261 fl.fl6_flowlabel = 0;
262 fl.oif = ireq6->iif;
263 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
264 fl.fl_ip_sport = inet_rsk(req)->loc_port;
265 security_req_classify_flow(req, &fl);
267 opt = np->opt;
269 if (opt != NULL && opt->srcrt != NULL) {
270 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
272 ipv6_addr_copy(&final, &fl.fl6_dst);
273 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
274 final_p = &final;
277 err = ip6_dst_lookup(sk, &dst, &fl);
278 if (err)
279 goto done;
281 if (final_p)
282 ipv6_addr_copy(&fl.fl6_dst, final_p);
284 err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0);
285 if (err < 0)
286 goto done;
288 skb = dccp_make_response(sk, dst, req);
289 if (skb != NULL) {
290 struct dccp_hdr *dh = dccp_hdr(skb);
292 dh->dccph_checksum = dccp_v6_csum_finish(skb,
293 &ireq6->loc_addr,
294 &ireq6->rmt_addr);
295 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
296 err = ip6_xmit(sk, skb, &fl, opt, 0);
297 err = net_xmit_eval(err);
300 done:
301 if (opt != NULL && opt != np->opt)
302 sock_kfree_s(sk, opt, opt->tot_len);
303 dst_release(dst);
304 return err;
307 static void dccp_v6_reqsk_destructor(struct request_sock *req)
309 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
310 if (inet6_rsk(req)->pktopts != NULL)
311 kfree_skb(inet6_rsk(req)->pktopts);
314 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
316 struct ipv6hdr *rxip6h;
317 struct sk_buff *skb;
318 struct flowi fl;
319 struct net *net = dev_net(skb_dst(rxskb)->dev);
320 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
321 struct dst_entry *dst;
323 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
324 return;
326 if (!ipv6_unicast_destination(rxskb))
327 return;
329 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
330 if (skb == NULL)
331 return;
333 rxip6h = ipv6_hdr(rxskb);
334 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
335 &rxip6h->daddr);
337 memset(&fl, 0, sizeof(fl));
338 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
339 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
341 fl.proto = IPPROTO_DCCP;
342 fl.oif = inet6_iif(rxskb);
343 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
344 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
345 security_skb_classify_flow(rxskb, &fl);
347 /* sk = NULL, but it is safe for now. RST socket required. */
348 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) {
349 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) {
350 skb_dst_set(skb, dst);
351 ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
352 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
353 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
354 return;
358 kfree_skb(skb);
361 static struct request_sock_ops dccp6_request_sock_ops = {
362 .family = AF_INET6,
363 .obj_size = sizeof(struct dccp6_request_sock),
364 .rtx_syn_ack = dccp_v6_send_response,
365 .send_ack = dccp_reqsk_send_ack,
366 .destructor = dccp_v6_reqsk_destructor,
367 .send_reset = dccp_v6_ctl_send_reset,
370 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
372 const struct dccp_hdr *dh = dccp_hdr(skb);
373 const struct ipv6hdr *iph = ipv6_hdr(skb);
374 struct sock *nsk;
375 struct request_sock **prev;
376 /* Find possible connection requests. */
377 struct request_sock *req = inet6_csk_search_req(sk, &prev,
378 dh->dccph_sport,
379 &iph->saddr,
380 &iph->daddr,
381 inet6_iif(skb));
382 if (req != NULL)
383 return dccp_check_req(sk, skb, req, prev);
385 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
386 &iph->saddr, dh->dccph_sport,
387 &iph->daddr, ntohs(dh->dccph_dport),
388 inet6_iif(skb));
389 if (nsk != NULL) {
390 if (nsk->sk_state != DCCP_TIME_WAIT) {
391 bh_lock_sock(nsk);
392 return nsk;
394 inet_twsk_put(inet_twsk(nsk));
395 return NULL;
398 return sk;
401 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
403 struct request_sock *req;
404 struct dccp_request_sock *dreq;
405 struct inet6_request_sock *ireq6;
406 struct ipv6_pinfo *np = inet6_sk(sk);
407 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
408 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
410 if (skb->protocol == htons(ETH_P_IP))
411 return dccp_v4_conn_request(sk, skb);
413 if (!ipv6_unicast_destination(skb))
414 return 0; /* discard, don't send a reset here */
416 if (dccp_bad_service_code(sk, service)) {
417 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
418 goto drop;
421 * There are no SYN attacks on IPv6, yet...
423 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
424 if (inet_csk_reqsk_queue_is_full(sk))
425 goto drop;
427 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
428 goto drop;
430 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
431 if (req == NULL)
432 goto drop;
434 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
435 goto drop_and_free;
437 dreq = dccp_rsk(req);
438 if (dccp_parse_options(sk, dreq, skb))
439 goto drop_and_free;
441 if (security_inet_conn_request(sk, skb, req))
442 goto drop_and_free;
444 ireq6 = inet6_rsk(req);
445 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
446 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
448 if (ipv6_opt_accepted(sk, skb) ||
449 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
450 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
451 atomic_inc(&skb->users);
452 ireq6->pktopts = skb;
454 ireq6->iif = sk->sk_bound_dev_if;
456 /* So that link locals have meaning */
457 if (!sk->sk_bound_dev_if &&
458 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
459 ireq6->iif = inet6_iif(skb);
462 * Step 3: Process LISTEN state
464 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
466 * In fact we defer setting S.GSR, S.SWL, S.SWH to
467 * dccp_create_openreq_child.
469 dreq->dreq_isr = dcb->dccpd_seq;
470 dreq->dreq_iss = dccp_v6_init_sequence(skb);
471 dreq->dreq_service = service;
473 if (dccp_v6_send_response(sk, req, NULL))
474 goto drop_and_free;
476 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
477 return 0;
479 drop_and_free:
480 reqsk_free(req);
481 drop:
482 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
483 return -1;
486 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
487 struct sk_buff *skb,
488 struct request_sock *req,
489 struct dst_entry *dst)
491 struct inet6_request_sock *ireq6 = inet6_rsk(req);
492 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
493 struct inet_sock *newinet;
494 struct dccp_sock *newdp;
495 struct dccp6_sock *newdp6;
496 struct sock *newsk;
497 struct ipv6_txoptions *opt;
499 if (skb->protocol == htons(ETH_P_IP)) {
501 * v6 mapped
503 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
504 if (newsk == NULL)
505 return NULL;
507 newdp6 = (struct dccp6_sock *)newsk;
508 newdp = dccp_sk(newsk);
509 newinet = inet_sk(newsk);
510 newinet->pinet6 = &newdp6->inet6;
511 newnp = inet6_sk(newsk);
513 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
515 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
517 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->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->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
646 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
648 __inet6_hash(newsk, NULL);
649 __inet_inherit_port(sk, newsk);
651 return newsk;
653 out_overflow:
654 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
655 out:
656 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
657 if (opt != NULL && opt != np->opt)
658 sock_kfree_s(sk, opt, opt->tot_len);
659 dst_release(dst);
660 return NULL;
663 /* The socket must have it's spinlock held when we get
664 * here.
666 * We have a potential double-lock case here, so even when
667 * doing backlog processing we use the BH locking scheme.
668 * This is because we cannot sleep with the original spinlock
669 * held.
671 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
673 struct ipv6_pinfo *np = inet6_sk(sk);
674 struct sk_buff *opt_skb = NULL;
676 /* Imagine: socket is IPv6. IPv4 packet arrives,
677 goes to IPv4 receive handler and backlogged.
678 From backlog it always goes here. Kerboom...
679 Fortunately, dccp_rcv_established and rcv_established
680 handle them correctly, but it is not case with
681 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
684 if (skb->protocol == htons(ETH_P_IP))
685 return dccp_v4_do_rcv(sk, skb);
687 if (sk_filter(sk, skb))
688 goto discard;
691 * socket locking is here for SMP purposes as backlog rcv is currently
692 * called with bh processing disabled.
695 /* Do Stevens' IPV6_PKTOPTIONS.
697 Yes, guys, it is the only place in our code, where we
698 may make it not affecting IPv4.
699 The rest of code is protocol independent,
700 and I do not like idea to uglify IPv4.
702 Actually, all the idea behind IPV6_PKTOPTIONS
703 looks not very well thought. For now we latch
704 options, received in the last packet, enqueued
705 by tcp. Feel free to propose better solution.
706 --ANK (980728)
708 if (np->rxopt.all)
710 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
711 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
713 opt_skb = skb_clone(skb, GFP_ATOMIC);
715 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
716 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
717 goto reset;
718 if (opt_skb) {
719 /* XXX This is where we would goto ipv6_pktoptions. */
720 __kfree_skb(opt_skb);
722 return 0;
726 * Step 3: Process LISTEN state
727 * If S.state == LISTEN,
728 * If P.type == Request or P contains a valid Init Cookie option,
729 * (* Must scan the packet's options to check for Init
730 * Cookies. Only Init Cookies are processed here,
731 * however; other options are processed in Step 8. This
732 * scan need only be performed if the endpoint uses Init
733 * Cookies *)
734 * (* Generate a new socket and switch to that socket *)
735 * Set S := new socket for this port pair
736 * S.state = RESPOND
737 * Choose S.ISS (initial seqno) or set from Init Cookies
738 * Initialize S.GAR := S.ISS
739 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
740 * Continue with S.state == RESPOND
741 * (* A Response packet will be generated in Step 11 *)
742 * Otherwise,
743 * Generate Reset(No Connection) unless P.type == Reset
744 * Drop packet and return
746 * NOTE: the check for the packet types is done in
747 * dccp_rcv_state_process
749 if (sk->sk_state == DCCP_LISTEN) {
750 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
752 if (nsk == NULL)
753 goto discard;
755 * Queue it on the new socket if the new socket is active,
756 * otherwise we just shortcircuit this and continue with
757 * the new socket..
759 if (nsk != sk) {
760 if (dccp_child_process(sk, nsk, skb))
761 goto reset;
762 if (opt_skb != NULL)
763 __kfree_skb(opt_skb);
764 return 0;
768 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
769 goto reset;
770 if (opt_skb) {
771 /* XXX This is where we would goto ipv6_pktoptions. */
772 __kfree_skb(opt_skb);
774 return 0;
776 reset:
777 dccp_v6_ctl_send_reset(sk, skb);
778 discard:
779 if (opt_skb != NULL)
780 __kfree_skb(opt_skb);
781 kfree_skb(skb);
782 return 0;
785 static int dccp_v6_rcv(struct sk_buff *skb)
787 const struct dccp_hdr *dh;
788 struct sock *sk;
789 int min_cov;
791 /* Step 1: Check header basics */
793 if (dccp_invalid_packet(skb))
794 goto discard_it;
796 /* Step 1: If header checksum is incorrect, drop packet and return. */
797 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
798 &ipv6_hdr(skb)->daddr)) {
799 DCCP_WARN("dropped packet with invalid checksum\n");
800 goto discard_it;
803 dh = dccp_hdr(skb);
805 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
806 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
808 if (dccp_packet_without_ack(skb))
809 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
810 else
811 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
813 /* Step 2:
814 * Look up flow ID in table and get corresponding socket */
815 sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
816 dh->dccph_sport, dh->dccph_dport);
818 * Step 2:
819 * If no socket ...
821 if (sk == NULL) {
822 dccp_pr_debug("failed to look up flow ID in table and "
823 "get corresponding socket\n");
824 goto no_dccp_socket;
828 * Step 2:
829 * ... or S.state == TIMEWAIT,
830 * Generate Reset(No Connection) unless P.type == Reset
831 * Drop packet and return
833 if (sk->sk_state == DCCP_TIME_WAIT) {
834 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
835 inet_twsk_put(inet_twsk(sk));
836 goto no_dccp_socket;
840 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
841 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
842 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
844 min_cov = dccp_sk(sk)->dccps_pcrlen;
845 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
846 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
847 dh->dccph_cscov, min_cov);
848 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
849 goto discard_and_relse;
852 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
853 goto discard_and_relse;
855 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
857 no_dccp_socket:
858 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
859 goto discard_it;
861 * Step 2:
862 * If no socket ...
863 * Generate Reset(No Connection) unless P.type == Reset
864 * Drop packet and return
866 if (dh->dccph_type != DCCP_PKT_RESET) {
867 DCCP_SKB_CB(skb)->dccpd_reset_code =
868 DCCP_RESET_CODE_NO_CONNECTION;
869 dccp_v6_ctl_send_reset(sk, skb);
872 discard_it:
873 kfree_skb(skb);
874 return 0;
876 discard_and_relse:
877 sock_put(sk);
878 goto discard_it;
881 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
882 int addr_len)
884 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
885 struct inet_connection_sock *icsk = inet_csk(sk);
886 struct inet_sock *inet = inet_sk(sk);
887 struct ipv6_pinfo *np = inet6_sk(sk);
888 struct dccp_sock *dp = dccp_sk(sk);
889 struct in6_addr *saddr = NULL, *final_p = NULL, final;
890 struct flowi fl;
891 struct dst_entry *dst;
892 int addr_type;
893 int err;
895 dp->dccps_role = DCCP_ROLE_CLIENT;
897 if (addr_len < SIN6_LEN_RFC2133)
898 return -EINVAL;
900 if (usin->sin6_family != AF_INET6)
901 return -EAFNOSUPPORT;
903 memset(&fl, 0, sizeof(fl));
905 if (np->sndflow) {
906 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
907 IP6_ECN_flow_init(fl.fl6_flowlabel);
908 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
909 struct ip6_flowlabel *flowlabel;
910 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
911 if (flowlabel == NULL)
912 return -EINVAL;
913 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
914 fl6_sock_release(flowlabel);
918 * connect() to INADDR_ANY means loopback (BSD'ism).
920 if (ipv6_addr_any(&usin->sin6_addr))
921 usin->sin6_addr.s6_addr[15] = 1;
923 addr_type = ipv6_addr_type(&usin->sin6_addr);
925 if (addr_type & IPV6_ADDR_MULTICAST)
926 return -ENETUNREACH;
928 if (addr_type & IPV6_ADDR_LINKLOCAL) {
929 if (addr_len >= sizeof(struct sockaddr_in6) &&
930 usin->sin6_scope_id) {
931 /* If interface is set while binding, indices
932 * must coincide.
934 if (sk->sk_bound_dev_if &&
935 sk->sk_bound_dev_if != usin->sin6_scope_id)
936 return -EINVAL;
938 sk->sk_bound_dev_if = usin->sin6_scope_id;
941 /* Connect to link-local address requires an interface */
942 if (!sk->sk_bound_dev_if)
943 return -EINVAL;
946 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
947 np->flow_label = fl.fl6_flowlabel;
950 * DCCP over IPv4
952 if (addr_type == IPV6_ADDR_MAPPED) {
953 u32 exthdrlen = icsk->icsk_ext_hdr_len;
954 struct sockaddr_in sin;
956 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
958 if (__ipv6_only_sock(sk))
959 return -ENETUNREACH;
961 sin.sin_family = AF_INET;
962 sin.sin_port = usin->sin6_port;
963 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
965 icsk->icsk_af_ops = &dccp_ipv6_mapped;
966 sk->sk_backlog_rcv = dccp_v4_do_rcv;
968 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
969 if (err) {
970 icsk->icsk_ext_hdr_len = exthdrlen;
971 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
972 sk->sk_backlog_rcv = dccp_v6_do_rcv;
973 goto failure;
975 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
976 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr);
978 return err;
981 if (!ipv6_addr_any(&np->rcv_saddr))
982 saddr = &np->rcv_saddr;
984 fl.proto = IPPROTO_DCCP;
985 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
986 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
987 fl.oif = sk->sk_bound_dev_if;
988 fl.fl_ip_dport = usin->sin6_port;
989 fl.fl_ip_sport = inet->inet_sport;
990 security_sk_classify_flow(sk, &fl);
992 if (np->opt != NULL && np->opt->srcrt != NULL) {
993 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
995 ipv6_addr_copy(&final, &fl.fl6_dst);
996 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
997 final_p = &final;
1000 err = ip6_dst_lookup(sk, &dst, &fl);
1001 if (err)
1002 goto failure;
1004 if (final_p)
1005 ipv6_addr_copy(&fl.fl6_dst, final_p);
1007 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
1008 if (err < 0) {
1009 if (err == -EREMOTE)
1010 err = ip6_dst_blackhole(sk, &dst, &fl);
1011 if (err < 0)
1012 goto failure;
1015 if (saddr == NULL) {
1016 saddr = &fl.fl6_src;
1017 ipv6_addr_copy(&np->rcv_saddr, saddr);
1020 /* set the source address */
1021 ipv6_addr_copy(&np->saddr, saddr);
1022 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
1024 __ip6_dst_store(sk, dst, NULL, NULL);
1026 icsk->icsk_ext_hdr_len = 0;
1027 if (np->opt != NULL)
1028 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1029 np->opt->opt_nflen);
1031 inet->inet_dport = usin->sin6_port;
1033 dccp_set_state(sk, DCCP_REQUESTING);
1034 err = inet6_hash_connect(&dccp_death_row, sk);
1035 if (err)
1036 goto late_failure;
1038 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1039 np->daddr.s6_addr32,
1040 inet->inet_sport,
1041 inet->inet_dport);
1042 err = dccp_connect(sk);
1043 if (err)
1044 goto late_failure;
1046 return 0;
1048 late_failure:
1049 dccp_set_state(sk, DCCP_CLOSED);
1050 __sk_dst_reset(sk);
1051 failure:
1052 inet->inet_dport = 0;
1053 sk->sk_route_caps = 0;
1054 return err;
1057 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1058 .queue_xmit = inet6_csk_xmit,
1059 .send_check = dccp_v6_send_check,
1060 .rebuild_header = inet6_sk_rebuild_header,
1061 .conn_request = dccp_v6_conn_request,
1062 .syn_recv_sock = dccp_v6_request_recv_sock,
1063 .net_header_len = sizeof(struct ipv6hdr),
1064 .setsockopt = ipv6_setsockopt,
1065 .getsockopt = ipv6_getsockopt,
1066 .addr2sockaddr = inet6_csk_addr2sockaddr,
1067 .sockaddr_len = sizeof(struct sockaddr_in6),
1068 .bind_conflict = inet6_csk_bind_conflict,
1069 #ifdef CONFIG_COMPAT
1070 .compat_setsockopt = compat_ipv6_setsockopt,
1071 .compat_getsockopt = compat_ipv6_getsockopt,
1072 #endif
1076 * DCCP over IPv4 via INET6 API
1078 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1079 .queue_xmit = ip_queue_xmit,
1080 .send_check = dccp_v4_send_check,
1081 .rebuild_header = inet_sk_rebuild_header,
1082 .conn_request = dccp_v6_conn_request,
1083 .syn_recv_sock = dccp_v6_request_recv_sock,
1084 .net_header_len = sizeof(struct iphdr),
1085 .setsockopt = ipv6_setsockopt,
1086 .getsockopt = ipv6_getsockopt,
1087 .addr2sockaddr = inet6_csk_addr2sockaddr,
1088 .sockaddr_len = sizeof(struct sockaddr_in6),
1089 #ifdef CONFIG_COMPAT
1090 .compat_setsockopt = compat_ipv6_setsockopt,
1091 .compat_getsockopt = compat_ipv6_getsockopt,
1092 #endif
1095 /* NOTE: A lot of things set to zero explicitly by call to
1096 * sk_alloc() so need not be done here.
1098 static int dccp_v6_init_sock(struct sock *sk)
1100 static __u8 dccp_v6_ctl_sock_initialized;
1101 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1103 if (err == 0) {
1104 if (unlikely(!dccp_v6_ctl_sock_initialized))
1105 dccp_v6_ctl_sock_initialized = 1;
1106 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1109 return err;
1112 static void dccp_v6_destroy_sock(struct sock *sk)
1114 dccp_destroy_sock(sk);
1115 inet6_destroy_sock(sk);
1118 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1119 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1122 static struct proto dccp_v6_prot = {
1123 .name = "DCCPv6",
1124 .owner = THIS_MODULE,
1125 .close = dccp_close,
1126 .connect = dccp_v6_connect,
1127 .disconnect = dccp_disconnect,
1128 .ioctl = dccp_ioctl,
1129 .init = dccp_v6_init_sock,
1130 .setsockopt = dccp_setsockopt,
1131 .getsockopt = dccp_getsockopt,
1132 .sendmsg = dccp_sendmsg,
1133 .recvmsg = dccp_recvmsg,
1134 .backlog_rcv = dccp_v6_do_rcv,
1135 .hash = dccp_v6_hash,
1136 .unhash = inet_unhash,
1137 .accept = inet_csk_accept,
1138 .get_port = inet_csk_get_port,
1139 .shutdown = dccp_shutdown,
1140 .destroy = dccp_v6_destroy_sock,
1141 .orphan_count = &dccp_orphan_count,
1142 .max_header = MAX_DCCP_HEADER,
1143 .obj_size = sizeof(struct dccp6_sock),
1144 .slab_flags = SLAB_DESTROY_BY_RCU,
1145 .rsk_prot = &dccp6_request_sock_ops,
1146 .twsk_prot = &dccp6_timewait_sock_ops,
1147 .h.hashinfo = &dccp_hashinfo,
1148 #ifdef CONFIG_COMPAT
1149 .compat_setsockopt = compat_dccp_setsockopt,
1150 .compat_getsockopt = compat_dccp_getsockopt,
1151 #endif
1154 static const struct inet6_protocol dccp_v6_protocol = {
1155 .handler = dccp_v6_rcv,
1156 .err_handler = dccp_v6_err,
1157 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1160 static const struct proto_ops inet6_dccp_ops = {
1161 .family = PF_INET6,
1162 .owner = THIS_MODULE,
1163 .release = inet6_release,
1164 .bind = inet6_bind,
1165 .connect = inet_stream_connect,
1166 .socketpair = sock_no_socketpair,
1167 .accept = inet_accept,
1168 .getname = inet6_getname,
1169 .poll = dccp_poll,
1170 .ioctl = inet6_ioctl,
1171 .listen = inet_dccp_listen,
1172 .shutdown = inet_shutdown,
1173 .setsockopt = sock_common_setsockopt,
1174 .getsockopt = sock_common_getsockopt,
1175 .sendmsg = inet_sendmsg,
1176 .recvmsg = sock_common_recvmsg,
1177 .mmap = sock_no_mmap,
1178 .sendpage = sock_no_sendpage,
1179 #ifdef CONFIG_COMPAT
1180 .compat_setsockopt = compat_sock_common_setsockopt,
1181 .compat_getsockopt = compat_sock_common_getsockopt,
1182 #endif
1185 static struct inet_protosw dccp_v6_protosw = {
1186 .type = SOCK_DCCP,
1187 .protocol = IPPROTO_DCCP,
1188 .prot = &dccp_v6_prot,
1189 .ops = &inet6_dccp_ops,
1190 .flags = INET_PROTOSW_ICSK,
1193 static int __net_init dccp_v6_init_net(struct net *net)
1195 if (dccp_hashinfo.bhash == NULL)
1196 return -ESOCKTNOSUPPORT;
1198 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1199 SOCK_DCCP, IPPROTO_DCCP, net);
1202 static void __net_exit dccp_v6_exit_net(struct net *net)
1204 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1207 static struct pernet_operations dccp_v6_ops = {
1208 .init = dccp_v6_init_net,
1209 .exit = dccp_v6_exit_net,
1212 static int __init dccp_v6_init(void)
1214 int err = proto_register(&dccp_v6_prot, 1);
1216 if (err != 0)
1217 goto out;
1219 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1220 if (err != 0)
1221 goto out_unregister_proto;
1223 inet6_register_protosw(&dccp_v6_protosw);
1225 err = register_pernet_subsys(&dccp_v6_ops);
1226 if (err != 0)
1227 goto out_destroy_ctl_sock;
1228 out:
1229 return err;
1231 out_destroy_ctl_sock:
1232 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1233 inet6_unregister_protosw(&dccp_v6_protosw);
1234 out_unregister_proto:
1235 proto_unregister(&dccp_v6_prot);
1236 goto out;
1239 static void __exit dccp_v6_exit(void)
1241 unregister_pernet_subsys(&dccp_v6_ops);
1242 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1243 inet6_unregister_protosw(&dccp_v6_protosw);
1244 proto_unregister(&dccp_v6_prot);
1247 module_init(dccp_v6_init);
1248 module_exit(dccp_v6_exit);
1251 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1252 * values directly, Also cover the case where the protocol is not specified,
1253 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1255 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1256 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1257 MODULE_LICENSE("GPL");
1258 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1259 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");