powerpc: Don't use kernel stack with translation off
[linux/fpc-iii.git] / net / dccp / ipv6.c
blob5e1ee0da2c40cab9c7f44ffd431e4df4455e08f5
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 struct inet_connection_sock_af_ops dccp_ipv6_mapped;
39 static 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 int type, int 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(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
102 return;
105 sk = inet6_lookup(net, &dccp_hashinfo,
106 &hdr->daddr, dh->dccph_dport,
107 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
109 if (sk == NULL) {
110 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
111 return;
114 if (sk->sk_state == DCCP_TIME_WAIT) {
115 inet_twsk_put(inet_twsk(sk));
116 return;
119 bh_lock_sock(sk);
120 if (sock_owned_by_user(sk))
121 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
123 if (sk->sk_state == DCCP_CLOSED)
124 goto out;
126 dp = dccp_sk(sk);
127 seq = dccp_hdr_seq(dh);
128 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
129 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
130 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
131 goto out;
134 np = inet6_sk(sk);
136 if (type == ICMPV6_PKT_TOOBIG) {
137 struct dst_entry *dst = NULL;
139 if (sock_owned_by_user(sk))
140 goto out;
141 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
142 goto out;
144 /* icmp should have updated the destination cache entry */
145 dst = __sk_dst_check(sk, np->dst_cookie);
146 if (dst == NULL) {
147 struct inet_sock *inet = inet_sk(sk);
148 struct flowi fl;
150 /* BUGGG_FUTURE: Again, it is not clear how
151 to handle rthdr case. Ignore this complexity
152 for now.
154 memset(&fl, 0, sizeof(fl));
155 fl.proto = IPPROTO_DCCP;
156 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
157 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
158 fl.oif = sk->sk_bound_dev_if;
159 fl.fl_ip_dport = inet->dport;
160 fl.fl_ip_sport = inet->sport;
161 security_sk_classify_flow(sk, &fl);
163 err = ip6_dst_lookup(sk, &dst, &fl);
164 if (err) {
165 sk->sk_err_soft = -err;
166 goto out;
169 err = xfrm_lookup(&dst, &fl, sk, 0);
170 if (err < 0) {
171 sk->sk_err_soft = -err;
172 goto out;
174 } else
175 dst_hold(dst);
177 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
178 dccp_sync_mss(sk, dst_mtu(dst));
179 } /* else let the usual retransmit timer handle it */
180 dst_release(dst);
181 goto out;
184 icmpv6_err_convert(type, code, &err);
186 /* Might be for an request_sock */
187 switch (sk->sk_state) {
188 struct request_sock *req, **prev;
189 case DCCP_LISTEN:
190 if (sock_owned_by_user(sk))
191 goto out;
193 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
194 &hdr->daddr, &hdr->saddr,
195 inet6_iif(skb));
196 if (req == NULL)
197 goto out;
200 * ICMPs are not backlogged, hence we cannot get an established
201 * socket here.
203 WARN_ON(req->sk != NULL);
205 if (seq != dccp_rsk(req)->dreq_iss) {
206 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
207 goto out;
210 inet_csk_reqsk_queue_drop(sk, req, prev);
211 goto out;
213 case DCCP_REQUESTING:
214 case DCCP_RESPOND: /* Cannot happen.
215 It can, it SYNs are crossed. --ANK */
216 if (!sock_owned_by_user(sk)) {
217 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
218 sk->sk_err = err;
220 * Wake people up to see the error
221 * (see connect in sock.c)
223 sk->sk_error_report(sk);
224 dccp_done(sk);
225 } else
226 sk->sk_err_soft = err;
227 goto out;
230 if (!sock_owned_by_user(sk) && np->recverr) {
231 sk->sk_err = err;
232 sk->sk_error_report(sk);
233 } else
234 sk->sk_err_soft = err;
236 out:
237 bh_unlock_sock(sk);
238 sock_put(sk);
242 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req)
244 struct inet6_request_sock *ireq6 = inet6_rsk(req);
245 struct ipv6_pinfo *np = inet6_sk(sk);
246 struct sk_buff *skb;
247 struct ipv6_txoptions *opt = NULL;
248 struct in6_addr *final_p = NULL, final;
249 struct flowi fl;
250 int err = -1;
251 struct dst_entry *dst;
253 memset(&fl, 0, sizeof(fl));
254 fl.proto = IPPROTO_DCCP;
255 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
256 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
257 fl.fl6_flowlabel = 0;
258 fl.oif = ireq6->iif;
259 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
260 fl.fl_ip_sport = inet_sk(sk)->sport;
261 security_req_classify_flow(req, &fl);
263 opt = np->opt;
265 if (opt != NULL && opt->srcrt != NULL) {
266 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
268 ipv6_addr_copy(&final, &fl.fl6_dst);
269 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
270 final_p = &final;
273 err = ip6_dst_lookup(sk, &dst, &fl);
274 if (err)
275 goto done;
277 if (final_p)
278 ipv6_addr_copy(&fl.fl6_dst, final_p);
280 err = xfrm_lookup(&dst, &fl, sk, 0);
281 if (err < 0)
282 goto done;
284 skb = dccp_make_response(sk, dst, req);
285 if (skb != NULL) {
286 struct dccp_hdr *dh = dccp_hdr(skb);
288 dh->dccph_checksum = dccp_v6_csum_finish(skb,
289 &ireq6->loc_addr,
290 &ireq6->rmt_addr);
291 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
292 err = ip6_xmit(sk, skb, &fl, opt, 0);
293 err = net_xmit_eval(err);
296 done:
297 if (opt != NULL && opt != np->opt)
298 sock_kfree_s(sk, opt, opt->tot_len);
299 dst_release(dst);
300 return err;
303 static void dccp_v6_reqsk_destructor(struct request_sock *req)
305 if (inet6_rsk(req)->pktopts != NULL)
306 kfree_skb(inet6_rsk(req)->pktopts);
309 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
311 struct ipv6hdr *rxip6h;
312 struct sk_buff *skb;
313 struct flowi fl;
314 struct net *net = dev_net(rxskb->dst->dev);
315 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
317 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
318 return;
320 if (!ipv6_unicast_destination(rxskb))
321 return;
323 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
324 if (skb == NULL)
325 return;
327 rxip6h = ipv6_hdr(rxskb);
328 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
329 &rxip6h->daddr);
331 memset(&fl, 0, sizeof(fl));
332 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
333 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
335 fl.proto = IPPROTO_DCCP;
336 fl.oif = inet6_iif(rxskb);
337 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
338 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
339 security_skb_classify_flow(rxskb, &fl);
341 /* sk = NULL, but it is safe for now. RST socket required. */
342 if (!ip6_dst_lookup(ctl_sk, &skb->dst, &fl)) {
343 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
344 ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
345 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
346 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
347 return;
351 kfree_skb(skb);
354 static struct request_sock_ops dccp6_request_sock_ops = {
355 .family = AF_INET6,
356 .obj_size = sizeof(struct dccp6_request_sock),
357 .rtx_syn_ack = dccp_v6_send_response,
358 .send_ack = dccp_reqsk_send_ack,
359 .destructor = dccp_v6_reqsk_destructor,
360 .send_reset = dccp_v6_ctl_send_reset,
363 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
365 const struct dccp_hdr *dh = dccp_hdr(skb);
366 const struct ipv6hdr *iph = ipv6_hdr(skb);
367 struct sock *nsk;
368 struct request_sock **prev;
369 /* Find possible connection requests. */
370 struct request_sock *req = inet6_csk_search_req(sk, &prev,
371 dh->dccph_sport,
372 &iph->saddr,
373 &iph->daddr,
374 inet6_iif(skb));
375 if (req != NULL)
376 return dccp_check_req(sk, skb, req, prev);
378 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
379 &iph->saddr, dh->dccph_sport,
380 &iph->daddr, ntohs(dh->dccph_dport),
381 inet6_iif(skb));
382 if (nsk != NULL) {
383 if (nsk->sk_state != DCCP_TIME_WAIT) {
384 bh_lock_sock(nsk);
385 return nsk;
387 inet_twsk_put(inet_twsk(nsk));
388 return NULL;
391 return sk;
394 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
396 struct request_sock *req;
397 struct dccp_request_sock *dreq;
398 struct inet6_request_sock *ireq6;
399 struct ipv6_pinfo *np = inet6_sk(sk);
400 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
401 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
403 if (skb->protocol == htons(ETH_P_IP))
404 return dccp_v4_conn_request(sk, skb);
406 if (!ipv6_unicast_destination(skb))
407 return 0; /* discard, don't send a reset here */
409 if (dccp_bad_service_code(sk, service)) {
410 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
411 goto drop;
414 * There are no SYN attacks on IPv6, yet...
416 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
417 if (inet_csk_reqsk_queue_is_full(sk))
418 goto drop;
420 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
421 goto drop;
423 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
424 if (req == NULL)
425 goto drop;
427 dccp_reqsk_init(req, skb);
429 dreq = dccp_rsk(req);
430 if (dccp_parse_options(sk, dreq, skb))
431 goto drop_and_free;
433 if (security_inet_conn_request(sk, skb, req))
434 goto drop_and_free;
436 ireq6 = inet6_rsk(req);
437 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
438 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
440 if (ipv6_opt_accepted(sk, skb) ||
441 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
442 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
443 atomic_inc(&skb->users);
444 ireq6->pktopts = skb;
446 ireq6->iif = sk->sk_bound_dev_if;
448 /* So that link locals have meaning */
449 if (!sk->sk_bound_dev_if &&
450 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
451 ireq6->iif = inet6_iif(skb);
454 * Step 3: Process LISTEN state
456 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
458 * In fact we defer setting S.GSR, S.SWL, S.SWH to
459 * dccp_create_openreq_child.
461 dreq->dreq_isr = dcb->dccpd_seq;
462 dreq->dreq_iss = dccp_v6_init_sequence(skb);
463 dreq->dreq_service = service;
465 if (dccp_v6_send_response(sk, req))
466 goto drop_and_free;
468 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
469 return 0;
471 drop_and_free:
472 reqsk_free(req);
473 drop:
474 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
475 return -1;
478 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
479 struct sk_buff *skb,
480 struct request_sock *req,
481 struct dst_entry *dst)
483 struct inet6_request_sock *ireq6 = inet6_rsk(req);
484 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
485 struct inet_sock *newinet;
486 struct dccp_sock *newdp;
487 struct dccp6_sock *newdp6;
488 struct sock *newsk;
489 struct ipv6_txoptions *opt;
491 if (skb->protocol == htons(ETH_P_IP)) {
493 * v6 mapped
495 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
496 if (newsk == NULL)
497 return NULL;
499 newdp6 = (struct dccp6_sock *)newsk;
500 newdp = dccp_sk(newsk);
501 newinet = inet_sk(newsk);
502 newinet->pinet6 = &newdp6->inet6;
503 newnp = inet6_sk(newsk);
505 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
507 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
508 newinet->daddr);
510 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
511 newinet->saddr);
513 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
515 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
516 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
517 newnp->pktoptions = NULL;
518 newnp->opt = NULL;
519 newnp->mcast_oif = inet6_iif(skb);
520 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
523 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
524 * here, dccp_create_openreq_child now does this for us, see the comment in
525 * that function for the gory details. -acme
528 /* It is tricky place. Until this moment IPv4 tcp
529 worked with IPv6 icsk.icsk_af_ops.
530 Sync it now.
532 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
534 return newsk;
537 opt = np->opt;
539 if (sk_acceptq_is_full(sk))
540 goto out_overflow;
542 if (dst == NULL) {
543 struct in6_addr *final_p = NULL, final;
544 struct flowi fl;
546 memset(&fl, 0, sizeof(fl));
547 fl.proto = IPPROTO_DCCP;
548 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
549 if (opt != NULL && opt->srcrt != NULL) {
550 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
552 ipv6_addr_copy(&final, &fl.fl6_dst);
553 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
554 final_p = &final;
556 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
557 fl.oif = sk->sk_bound_dev_if;
558 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
559 fl.fl_ip_sport = inet_sk(sk)->sport;
560 security_sk_classify_flow(sk, &fl);
562 if (ip6_dst_lookup(sk, &dst, &fl))
563 goto out;
565 if (final_p)
566 ipv6_addr_copy(&fl.fl6_dst, final_p);
568 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
569 goto out;
572 newsk = dccp_create_openreq_child(sk, req, skb);
573 if (newsk == NULL)
574 goto out;
577 * No need to charge this sock to the relevant IPv6 refcnt debug socks
578 * count here, dccp_create_openreq_child now does this for us, see the
579 * comment in that function for the gory details. -acme
582 __ip6_dst_store(newsk, dst, NULL, NULL);
583 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
584 NETIF_F_TSO);
585 newdp6 = (struct dccp6_sock *)newsk;
586 newinet = inet_sk(newsk);
587 newinet->pinet6 = &newdp6->inet6;
588 newdp = dccp_sk(newsk);
589 newnp = inet6_sk(newsk);
591 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
593 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
594 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
595 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
596 newsk->sk_bound_dev_if = ireq6->iif;
598 /* Now IPv6 options...
600 First: no IPv4 options.
602 newinet->opt = NULL;
604 /* Clone RX bits */
605 newnp->rxopt.all = np->rxopt.all;
607 /* Clone pktoptions received with SYN */
608 newnp->pktoptions = NULL;
609 if (ireq6->pktopts != NULL) {
610 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
611 kfree_skb(ireq6->pktopts);
612 ireq6->pktopts = NULL;
613 if (newnp->pktoptions)
614 skb_set_owner_r(newnp->pktoptions, newsk);
616 newnp->opt = NULL;
617 newnp->mcast_oif = inet6_iif(skb);
618 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
621 * Clone native IPv6 options from listening socket (if any)
623 * Yes, keeping reference count would be much more clever, but we make
624 * one more one thing there: reattach optmem to newsk.
626 if (opt != NULL) {
627 newnp->opt = ipv6_dup_options(newsk, opt);
628 if (opt != np->opt)
629 sock_kfree_s(sk, opt, opt->tot_len);
632 inet_csk(newsk)->icsk_ext_hdr_len = 0;
633 if (newnp->opt != NULL)
634 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
635 newnp->opt->opt_flen);
637 dccp_sync_mss(newsk, dst_mtu(dst));
639 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
641 __inet6_hash(newsk);
642 __inet_inherit_port(sk, newsk);
644 return newsk;
646 out_overflow:
647 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
648 out:
649 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
650 if (opt != NULL && opt != np->opt)
651 sock_kfree_s(sk, opt, opt->tot_len);
652 dst_release(dst);
653 return NULL;
656 /* The socket must have it's spinlock held when we get
657 * here.
659 * We have a potential double-lock case here, so even when
660 * doing backlog processing we use the BH locking scheme.
661 * This is because we cannot sleep with the original spinlock
662 * held.
664 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
666 struct ipv6_pinfo *np = inet6_sk(sk);
667 struct sk_buff *opt_skb = NULL;
669 /* Imagine: socket is IPv6. IPv4 packet arrives,
670 goes to IPv4 receive handler and backlogged.
671 From backlog it always goes here. Kerboom...
672 Fortunately, dccp_rcv_established and rcv_established
673 handle them correctly, but it is not case with
674 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
677 if (skb->protocol == htons(ETH_P_IP))
678 return dccp_v4_do_rcv(sk, skb);
680 if (sk_filter(sk, skb))
681 goto discard;
684 * socket locking is here for SMP purposes as backlog rcv is currently
685 * called with bh processing disabled.
688 /* Do Stevens' IPV6_PKTOPTIONS.
690 Yes, guys, it is the only place in our code, where we
691 may make it not affecting IPv4.
692 The rest of code is protocol independent,
693 and I do not like idea to uglify IPv4.
695 Actually, all the idea behind IPV6_PKTOPTIONS
696 looks not very well thought. For now we latch
697 options, received in the last packet, enqueued
698 by tcp. Feel free to propose better solution.
699 --ANK (980728)
701 if (np->rxopt.all)
703 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
704 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
706 opt_skb = skb_clone(skb, GFP_ATOMIC);
708 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
709 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
710 goto reset;
711 if (opt_skb) {
712 /* XXX This is where we would goto ipv6_pktoptions. */
713 __kfree_skb(opt_skb);
715 return 0;
719 * Step 3: Process LISTEN state
720 * If S.state == LISTEN,
721 * If P.type == Request or P contains a valid Init Cookie option,
722 * (* Must scan the packet's options to check for Init
723 * Cookies. Only Init Cookies are processed here,
724 * however; other options are processed in Step 8. This
725 * scan need only be performed if the endpoint uses Init
726 * Cookies *)
727 * (* Generate a new socket and switch to that socket *)
728 * Set S := new socket for this port pair
729 * S.state = RESPOND
730 * Choose S.ISS (initial seqno) or set from Init Cookies
731 * Initialize S.GAR := S.ISS
732 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
733 * Continue with S.state == RESPOND
734 * (* A Response packet will be generated in Step 11 *)
735 * Otherwise,
736 * Generate Reset(No Connection) unless P.type == Reset
737 * Drop packet and return
739 * NOTE: the check for the packet types is done in
740 * dccp_rcv_state_process
742 if (sk->sk_state == DCCP_LISTEN) {
743 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
745 if (nsk == NULL)
746 goto discard;
748 * Queue it on the new socket if the new socket is active,
749 * otherwise we just shortcircuit this and continue with
750 * the new socket..
752 if (nsk != sk) {
753 if (dccp_child_process(sk, nsk, skb))
754 goto reset;
755 if (opt_skb != NULL)
756 __kfree_skb(opt_skb);
757 return 0;
761 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
762 goto reset;
763 if (opt_skb) {
764 /* XXX This is where we would goto ipv6_pktoptions. */
765 __kfree_skb(opt_skb);
767 return 0;
769 reset:
770 dccp_v6_ctl_send_reset(sk, skb);
771 discard:
772 if (opt_skb != NULL)
773 __kfree_skb(opt_skb);
774 kfree_skb(skb);
775 return 0;
778 static int dccp_v6_rcv(struct sk_buff *skb)
780 const struct dccp_hdr *dh;
781 struct sock *sk;
782 int min_cov;
784 /* Step 1: Check header basics */
786 if (dccp_invalid_packet(skb))
787 goto discard_it;
789 /* Step 1: If header checksum is incorrect, drop packet and return. */
790 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
791 &ipv6_hdr(skb)->daddr)) {
792 DCCP_WARN("dropped packet with invalid checksum\n");
793 goto discard_it;
796 dh = dccp_hdr(skb);
798 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
799 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
801 if (dccp_packet_without_ack(skb))
802 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
803 else
804 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
806 /* Step 2:
807 * Look up flow ID in table and get corresponding socket */
808 sk = __inet6_lookup(dev_net(skb->dst->dev), &dccp_hashinfo,
809 &ipv6_hdr(skb)->saddr, dh->dccph_sport,
810 &ipv6_hdr(skb)->daddr, ntohs(dh->dccph_dport),
811 inet6_iif(skb));
813 * Step 2:
814 * If no socket ...
816 if (sk == NULL) {
817 dccp_pr_debug("failed to look up flow ID in table and "
818 "get corresponding socket\n");
819 goto no_dccp_socket;
823 * Step 2:
824 * ... or S.state == TIMEWAIT,
825 * Generate Reset(No Connection) unless P.type == Reset
826 * Drop packet and return
828 if (sk->sk_state == DCCP_TIME_WAIT) {
829 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
830 inet_twsk_put(inet_twsk(sk));
831 goto no_dccp_socket;
835 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
836 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
837 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
839 min_cov = dccp_sk(sk)->dccps_pcrlen;
840 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
841 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
842 dh->dccph_cscov, min_cov);
843 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
844 goto discard_and_relse;
847 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
848 goto discard_and_relse;
850 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
852 no_dccp_socket:
853 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
854 goto discard_it;
856 * Step 2:
857 * If no socket ...
858 * Generate Reset(No Connection) unless P.type == Reset
859 * Drop packet and return
861 if (dh->dccph_type != DCCP_PKT_RESET) {
862 DCCP_SKB_CB(skb)->dccpd_reset_code =
863 DCCP_RESET_CODE_NO_CONNECTION;
864 dccp_v6_ctl_send_reset(sk, skb);
867 discard_it:
868 kfree_skb(skb);
869 return 0;
871 discard_and_relse:
872 sock_put(sk);
873 goto discard_it;
876 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
877 int addr_len)
879 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
880 struct inet_connection_sock *icsk = inet_csk(sk);
881 struct inet_sock *inet = inet_sk(sk);
882 struct ipv6_pinfo *np = inet6_sk(sk);
883 struct dccp_sock *dp = dccp_sk(sk);
884 struct in6_addr *saddr = NULL, *final_p = NULL, final;
885 struct flowi fl;
886 struct dst_entry *dst;
887 int addr_type;
888 int err;
890 dp->dccps_role = DCCP_ROLE_CLIENT;
892 if (addr_len < SIN6_LEN_RFC2133)
893 return -EINVAL;
895 if (usin->sin6_family != AF_INET6)
896 return -EAFNOSUPPORT;
898 memset(&fl, 0, sizeof(fl));
900 if (np->sndflow) {
901 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
902 IP6_ECN_flow_init(fl.fl6_flowlabel);
903 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
904 struct ip6_flowlabel *flowlabel;
905 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
906 if (flowlabel == NULL)
907 return -EINVAL;
908 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
909 fl6_sock_release(flowlabel);
913 * connect() to INADDR_ANY means loopback (BSD'ism).
915 if (ipv6_addr_any(&usin->sin6_addr))
916 usin->sin6_addr.s6_addr[15] = 1;
918 addr_type = ipv6_addr_type(&usin->sin6_addr);
920 if (addr_type & IPV6_ADDR_MULTICAST)
921 return -ENETUNREACH;
923 if (addr_type & IPV6_ADDR_LINKLOCAL) {
924 if (addr_len >= sizeof(struct sockaddr_in6) &&
925 usin->sin6_scope_id) {
926 /* If interface is set while binding, indices
927 * must coincide.
929 if (sk->sk_bound_dev_if &&
930 sk->sk_bound_dev_if != usin->sin6_scope_id)
931 return -EINVAL;
933 sk->sk_bound_dev_if = usin->sin6_scope_id;
936 /* Connect to link-local address requires an interface */
937 if (!sk->sk_bound_dev_if)
938 return -EINVAL;
941 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
942 np->flow_label = fl.fl6_flowlabel;
945 * DCCP over IPv4
947 if (addr_type == IPV6_ADDR_MAPPED) {
948 u32 exthdrlen = icsk->icsk_ext_hdr_len;
949 struct sockaddr_in sin;
951 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
953 if (__ipv6_only_sock(sk))
954 return -ENETUNREACH;
956 sin.sin_family = AF_INET;
957 sin.sin_port = usin->sin6_port;
958 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
960 icsk->icsk_af_ops = &dccp_ipv6_mapped;
961 sk->sk_backlog_rcv = dccp_v4_do_rcv;
963 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
964 if (err) {
965 icsk->icsk_ext_hdr_len = exthdrlen;
966 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
967 sk->sk_backlog_rcv = dccp_v6_do_rcv;
968 goto failure;
969 } else {
970 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
971 inet->saddr);
972 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
973 inet->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(&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 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 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 .rsk_prot = &dccp6_request_sock_ops,
1142 .twsk_prot = &dccp6_timewait_sock_ops,
1143 .h.hashinfo = &dccp_hashinfo,
1144 #ifdef CONFIG_COMPAT
1145 .compat_setsockopt = compat_dccp_setsockopt,
1146 .compat_getsockopt = compat_dccp_getsockopt,
1147 #endif
1150 static struct inet6_protocol dccp_v6_protocol = {
1151 .handler = dccp_v6_rcv,
1152 .err_handler = dccp_v6_err,
1153 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1156 static struct proto_ops inet6_dccp_ops = {
1157 .family = PF_INET6,
1158 .owner = THIS_MODULE,
1159 .release = inet6_release,
1160 .bind = inet6_bind,
1161 .connect = inet_stream_connect,
1162 .socketpair = sock_no_socketpair,
1163 .accept = inet_accept,
1164 .getname = inet6_getname,
1165 .poll = dccp_poll,
1166 .ioctl = inet6_ioctl,
1167 .listen = inet_dccp_listen,
1168 .shutdown = inet_shutdown,
1169 .setsockopt = sock_common_setsockopt,
1170 .getsockopt = sock_common_getsockopt,
1171 .sendmsg = inet_sendmsg,
1172 .recvmsg = sock_common_recvmsg,
1173 .mmap = sock_no_mmap,
1174 .sendpage = sock_no_sendpage,
1175 #ifdef CONFIG_COMPAT
1176 .compat_setsockopt = compat_sock_common_setsockopt,
1177 .compat_getsockopt = compat_sock_common_getsockopt,
1178 #endif
1181 static struct inet_protosw dccp_v6_protosw = {
1182 .type = SOCK_DCCP,
1183 .protocol = IPPROTO_DCCP,
1184 .prot = &dccp_v6_prot,
1185 .ops = &inet6_dccp_ops,
1186 .capability = -1,
1187 .flags = INET_PROTOSW_ICSK,
1190 static int dccp_v6_init_net(struct net *net)
1192 int err;
1194 err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1195 SOCK_DCCP, IPPROTO_DCCP, net);
1196 return err;
1199 static void dccp_v6_exit_net(struct net *net)
1201 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1204 static struct pernet_operations dccp_v6_ops = {
1205 .init = dccp_v6_init_net,
1206 .exit = dccp_v6_exit_net,
1209 static int __init dccp_v6_init(void)
1211 int err = proto_register(&dccp_v6_prot, 1);
1213 if (err != 0)
1214 goto out;
1216 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1217 if (err != 0)
1218 goto out_unregister_proto;
1220 inet6_register_protosw(&dccp_v6_protosw);
1222 err = register_pernet_subsys(&dccp_v6_ops);
1223 if (err != 0)
1224 goto out_destroy_ctl_sock;
1225 out:
1226 return err;
1228 out_destroy_ctl_sock:
1229 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1230 inet6_unregister_protosw(&dccp_v6_protosw);
1231 out_unregister_proto:
1232 proto_unregister(&dccp_v6_prot);
1233 goto out;
1236 static void __exit dccp_v6_exit(void)
1238 unregister_pernet_subsys(&dccp_v6_ops);
1239 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1240 inet6_unregister_protosw(&dccp_v6_protosw);
1241 proto_unregister(&dccp_v6_prot);
1244 module_init(dccp_v6_init);
1245 module_exit(dccp_v6_exit);
1248 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1249 * values directly, Also cover the case where the protocol is not specified,
1250 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1252 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1253 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1254 MODULE_LICENSE("GPL");
1255 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1256 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");