nfsd: rename MAY_ flags
[linux-2.6/mini2440.git] / net / dccp / ipv6.c
blob9b1129bb7ece2c868316584d12805168dbbea017
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 ipv6_pinfo *np;
93 struct sock *sk;
94 int err;
95 __u64 seq;
97 sk = inet6_lookup(dev_net(skb->dev), &dccp_hashinfo,
98 &hdr->daddr, dh->dccph_dport,
99 &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
101 if (sk == NULL) {
102 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
103 return;
106 if (sk->sk_state == DCCP_TIME_WAIT) {
107 inet_twsk_put(inet_twsk(sk));
108 return;
111 bh_lock_sock(sk);
112 if (sock_owned_by_user(sk))
113 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
115 if (sk->sk_state == DCCP_CLOSED)
116 goto out;
118 np = inet6_sk(sk);
120 if (type == ICMPV6_PKT_TOOBIG) {
121 struct dst_entry *dst = NULL;
123 if (sock_owned_by_user(sk))
124 goto out;
125 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
126 goto out;
128 /* icmp should have updated the destination cache entry */
129 dst = __sk_dst_check(sk, np->dst_cookie);
130 if (dst == NULL) {
131 struct inet_sock *inet = inet_sk(sk);
132 struct flowi fl;
134 /* BUGGG_FUTURE: Again, it is not clear how
135 to handle rthdr case. Ignore this complexity
136 for now.
138 memset(&fl, 0, sizeof(fl));
139 fl.proto = IPPROTO_DCCP;
140 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
141 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
142 fl.oif = sk->sk_bound_dev_if;
143 fl.fl_ip_dport = inet->dport;
144 fl.fl_ip_sport = inet->sport;
145 security_sk_classify_flow(sk, &fl);
147 err = ip6_dst_lookup(sk, &dst, &fl);
148 if (err) {
149 sk->sk_err_soft = -err;
150 goto out;
153 err = xfrm_lookup(&dst, &fl, sk, 0);
154 if (err < 0) {
155 sk->sk_err_soft = -err;
156 goto out;
158 } else
159 dst_hold(dst);
161 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
162 dccp_sync_mss(sk, dst_mtu(dst));
163 } /* else let the usual retransmit timer handle it */
164 dst_release(dst);
165 goto out;
168 icmpv6_err_convert(type, code, &err);
170 seq = dccp_hdr_seq(dh);
171 /* Might be for an request_sock */
172 switch (sk->sk_state) {
173 struct request_sock *req, **prev;
174 case DCCP_LISTEN:
175 if (sock_owned_by_user(sk))
176 goto out;
178 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport,
179 &hdr->daddr, &hdr->saddr,
180 inet6_iif(skb));
181 if (req == NULL)
182 goto out;
185 * ICMPs are not backlogged, hence we cannot get an established
186 * socket here.
188 BUG_TRAP(req->sk == NULL);
190 if (seq != dccp_rsk(req)->dreq_iss) {
191 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
192 goto out;
195 inet_csk_reqsk_queue_drop(sk, req, prev);
196 goto out;
198 case DCCP_REQUESTING:
199 case DCCP_RESPOND: /* Cannot happen.
200 It can, it SYNs are crossed. --ANK */
201 if (!sock_owned_by_user(sk)) {
202 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
203 sk->sk_err = err;
205 * Wake people up to see the error
206 * (see connect in sock.c)
208 sk->sk_error_report(sk);
209 dccp_done(sk);
210 } else
211 sk->sk_err_soft = err;
212 goto out;
215 if (!sock_owned_by_user(sk) && np->recverr) {
216 sk->sk_err = err;
217 sk->sk_error_report(sk);
218 } else
219 sk->sk_err_soft = err;
221 out:
222 bh_unlock_sock(sk);
223 sock_put(sk);
227 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req)
229 struct inet6_request_sock *ireq6 = inet6_rsk(req);
230 struct ipv6_pinfo *np = inet6_sk(sk);
231 struct sk_buff *skb;
232 struct ipv6_txoptions *opt = NULL;
233 struct in6_addr *final_p = NULL, final;
234 struct flowi fl;
235 int err = -1;
236 struct dst_entry *dst;
238 memset(&fl, 0, sizeof(fl));
239 fl.proto = IPPROTO_DCCP;
240 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
241 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
242 fl.fl6_flowlabel = 0;
243 fl.oif = ireq6->iif;
244 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
245 fl.fl_ip_sport = inet_sk(sk)->sport;
246 security_req_classify_flow(req, &fl);
248 opt = np->opt;
250 if (opt != NULL && opt->srcrt != NULL) {
251 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
253 ipv6_addr_copy(&final, &fl.fl6_dst);
254 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
255 final_p = &final;
258 err = ip6_dst_lookup(sk, &dst, &fl);
259 if (err)
260 goto done;
262 if (final_p)
263 ipv6_addr_copy(&fl.fl6_dst, final_p);
265 err = xfrm_lookup(&dst, &fl, sk, 0);
266 if (err < 0)
267 goto done;
269 skb = dccp_make_response(sk, dst, req);
270 if (skb != NULL) {
271 struct dccp_hdr *dh = dccp_hdr(skb);
273 dh->dccph_checksum = dccp_v6_csum_finish(skb,
274 &ireq6->loc_addr,
275 &ireq6->rmt_addr);
276 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
277 err = ip6_xmit(sk, skb, &fl, opt, 0);
278 err = net_xmit_eval(err);
281 done:
282 if (opt != NULL && opt != np->opt)
283 sock_kfree_s(sk, opt, opt->tot_len);
284 dst_release(dst);
285 return err;
288 static void dccp_v6_reqsk_destructor(struct request_sock *req)
290 if (inet6_rsk(req)->pktopts != NULL)
291 kfree_skb(inet6_rsk(req)->pktopts);
294 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
296 struct ipv6hdr *rxip6h;
297 struct sk_buff *skb;
298 struct flowi fl;
299 struct net *net = dev_net(rxskb->dst->dev);
300 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
302 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
303 return;
305 if (!ipv6_unicast_destination(rxskb))
306 return;
308 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
309 if (skb == NULL)
310 return;
312 rxip6h = ipv6_hdr(rxskb);
313 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
314 &rxip6h->daddr);
316 memset(&fl, 0, sizeof(fl));
317 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
318 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
320 fl.proto = IPPROTO_DCCP;
321 fl.oif = inet6_iif(rxskb);
322 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
323 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
324 security_skb_classify_flow(rxskb, &fl);
326 /* sk = NULL, but it is safe for now. RST socket required. */
327 if (!ip6_dst_lookup(ctl_sk, &skb->dst, &fl)) {
328 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
329 ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
330 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
331 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
332 return;
336 kfree_skb(skb);
339 static struct request_sock_ops dccp6_request_sock_ops = {
340 .family = AF_INET6,
341 .obj_size = sizeof(struct dccp6_request_sock),
342 .rtx_syn_ack = dccp_v6_send_response,
343 .send_ack = dccp_reqsk_send_ack,
344 .destructor = dccp_v6_reqsk_destructor,
345 .send_reset = dccp_v6_ctl_send_reset,
348 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
350 const struct dccp_hdr *dh = dccp_hdr(skb);
351 const struct ipv6hdr *iph = ipv6_hdr(skb);
352 struct sock *nsk;
353 struct request_sock **prev;
354 /* Find possible connection requests. */
355 struct request_sock *req = inet6_csk_search_req(sk, &prev,
356 dh->dccph_sport,
357 &iph->saddr,
358 &iph->daddr,
359 inet6_iif(skb));
360 if (req != NULL)
361 return dccp_check_req(sk, skb, req, prev);
363 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
364 &iph->saddr, dh->dccph_sport,
365 &iph->daddr, ntohs(dh->dccph_dport),
366 inet6_iif(skb));
367 if (nsk != NULL) {
368 if (nsk->sk_state != DCCP_TIME_WAIT) {
369 bh_lock_sock(nsk);
370 return nsk;
372 inet_twsk_put(inet_twsk(nsk));
373 return NULL;
376 return sk;
379 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
381 struct request_sock *req;
382 struct dccp_request_sock *dreq;
383 struct inet6_request_sock *ireq6;
384 struct ipv6_pinfo *np = inet6_sk(sk);
385 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
386 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
388 if (skb->protocol == htons(ETH_P_IP))
389 return dccp_v4_conn_request(sk, skb);
391 if (!ipv6_unicast_destination(skb))
392 return 0; /* discard, don't send a reset here */
394 if (dccp_bad_service_code(sk, service)) {
395 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
396 goto drop;
399 * There are no SYN attacks on IPv6, yet...
401 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
402 if (inet_csk_reqsk_queue_is_full(sk))
403 goto drop;
405 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
406 goto drop;
408 req = inet6_reqsk_alloc(&dccp6_request_sock_ops);
409 if (req == NULL)
410 goto drop;
412 dccp_reqsk_init(req, skb);
414 dreq = dccp_rsk(req);
415 if (dccp_parse_options(sk, dreq, skb))
416 goto drop_and_free;
418 if (security_inet_conn_request(sk, skb, req))
419 goto drop_and_free;
421 ireq6 = inet6_rsk(req);
422 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
423 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
424 ireq6->pktopts = NULL;
426 if (ipv6_opt_accepted(sk, skb) ||
427 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
428 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
429 atomic_inc(&skb->users);
430 ireq6->pktopts = skb;
432 ireq6->iif = sk->sk_bound_dev_if;
434 /* So that link locals have meaning */
435 if (!sk->sk_bound_dev_if &&
436 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL)
437 ireq6->iif = inet6_iif(skb);
440 * Step 3: Process LISTEN state
442 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
444 * In fact we defer setting S.GSR, S.SWL, S.SWH to
445 * dccp_create_openreq_child.
447 dreq->dreq_isr = dcb->dccpd_seq;
448 dreq->dreq_iss = dccp_v6_init_sequence(skb);
449 dreq->dreq_service = service;
451 if (dccp_v6_send_response(sk, req))
452 goto drop_and_free;
454 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
455 return 0;
457 drop_and_free:
458 reqsk_free(req);
459 drop:
460 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
461 return -1;
464 static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
465 struct sk_buff *skb,
466 struct request_sock *req,
467 struct dst_entry *dst)
469 struct inet6_request_sock *ireq6 = inet6_rsk(req);
470 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
471 struct inet_sock *newinet;
472 struct dccp_sock *newdp;
473 struct dccp6_sock *newdp6;
474 struct sock *newsk;
475 struct ipv6_txoptions *opt;
477 if (skb->protocol == htons(ETH_P_IP)) {
479 * v6 mapped
481 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst);
482 if (newsk == NULL)
483 return NULL;
485 newdp6 = (struct dccp6_sock *)newsk;
486 newdp = dccp_sk(newsk);
487 newinet = inet_sk(newsk);
488 newinet->pinet6 = &newdp6->inet6;
489 newnp = inet6_sk(newsk);
491 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
493 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
494 newinet->daddr);
496 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
497 newinet->saddr);
499 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
501 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
502 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
503 newnp->pktoptions = NULL;
504 newnp->opt = NULL;
505 newnp->mcast_oif = inet6_iif(skb);
506 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
509 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
510 * here, dccp_create_openreq_child now does this for us, see the comment in
511 * that function for the gory details. -acme
514 /* It is tricky place. Until this moment IPv4 tcp
515 worked with IPv6 icsk.icsk_af_ops.
516 Sync it now.
518 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
520 return newsk;
523 opt = np->opt;
525 if (sk_acceptq_is_full(sk))
526 goto out_overflow;
528 if (dst == NULL) {
529 struct in6_addr *final_p = NULL, final;
530 struct flowi fl;
532 memset(&fl, 0, sizeof(fl));
533 fl.proto = IPPROTO_DCCP;
534 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr);
535 if (opt != NULL && opt->srcrt != NULL) {
536 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
538 ipv6_addr_copy(&final, &fl.fl6_dst);
539 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
540 final_p = &final;
542 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr);
543 fl.oif = sk->sk_bound_dev_if;
544 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
545 fl.fl_ip_sport = inet_sk(sk)->sport;
546 security_sk_classify_flow(sk, &fl);
548 if (ip6_dst_lookup(sk, &dst, &fl))
549 goto out;
551 if (final_p)
552 ipv6_addr_copy(&fl.fl6_dst, final_p);
554 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
555 goto out;
558 newsk = dccp_create_openreq_child(sk, req, skb);
559 if (newsk == NULL)
560 goto out;
563 * No need to charge this sock to the relevant IPv6 refcnt debug socks
564 * count here, dccp_create_openreq_child now does this for us, see the
565 * comment in that function for the gory details. -acme
568 __ip6_dst_store(newsk, dst, NULL, NULL);
569 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
570 NETIF_F_TSO);
571 newdp6 = (struct dccp6_sock *)newsk;
572 newinet = inet_sk(newsk);
573 newinet->pinet6 = &newdp6->inet6;
574 newdp = dccp_sk(newsk);
575 newnp = inet6_sk(newsk);
577 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
579 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr);
580 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr);
581 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr);
582 newsk->sk_bound_dev_if = ireq6->iif;
584 /* Now IPv6 options...
586 First: no IPv4 options.
588 newinet->opt = NULL;
590 /* Clone RX bits */
591 newnp->rxopt.all = np->rxopt.all;
593 /* Clone pktoptions received with SYN */
594 newnp->pktoptions = NULL;
595 if (ireq6->pktopts != NULL) {
596 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC);
597 kfree_skb(ireq6->pktopts);
598 ireq6->pktopts = NULL;
599 if (newnp->pktoptions)
600 skb_set_owner_r(newnp->pktoptions, newsk);
602 newnp->opt = NULL;
603 newnp->mcast_oif = inet6_iif(skb);
604 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
607 * Clone native IPv6 options from listening socket (if any)
609 * Yes, keeping reference count would be much more clever, but we make
610 * one more one thing there: reattach optmem to newsk.
612 if (opt != NULL) {
613 newnp->opt = ipv6_dup_options(newsk, opt);
614 if (opt != np->opt)
615 sock_kfree_s(sk, opt, opt->tot_len);
618 inet_csk(newsk)->icsk_ext_hdr_len = 0;
619 if (newnp->opt != NULL)
620 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
621 newnp->opt->opt_flen);
623 dccp_sync_mss(newsk, dst_mtu(dst));
625 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
627 __inet6_hash(newsk);
628 __inet_inherit_port(sk, newsk);
630 return newsk;
632 out_overflow:
633 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
634 out:
635 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
636 if (opt != NULL && opt != np->opt)
637 sock_kfree_s(sk, opt, opt->tot_len);
638 dst_release(dst);
639 return NULL;
642 /* The socket must have it's spinlock held when we get
643 * here.
645 * We have a potential double-lock case here, so even when
646 * doing backlog processing we use the BH locking scheme.
647 * This is because we cannot sleep with the original spinlock
648 * held.
650 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
652 struct ipv6_pinfo *np = inet6_sk(sk);
653 struct sk_buff *opt_skb = NULL;
655 /* Imagine: socket is IPv6. IPv4 packet arrives,
656 goes to IPv4 receive handler and backlogged.
657 From backlog it always goes here. Kerboom...
658 Fortunately, dccp_rcv_established and rcv_established
659 handle them correctly, but it is not case with
660 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
663 if (skb->protocol == htons(ETH_P_IP))
664 return dccp_v4_do_rcv(sk, skb);
666 if (sk_filter(sk, skb))
667 goto discard;
670 * socket locking is here for SMP purposes as backlog rcv is currently
671 * called with bh processing disabled.
674 /* Do Stevens' IPV6_PKTOPTIONS.
676 Yes, guys, it is the only place in our code, where we
677 may make it not affecting IPv4.
678 The rest of code is protocol independent,
679 and I do not like idea to uglify IPv4.
681 Actually, all the idea behind IPV6_PKTOPTIONS
682 looks not very well thought. For now we latch
683 options, received in the last packet, enqueued
684 by tcp. Feel free to propose better solution.
685 --ANK (980728)
687 if (np->rxopt.all)
689 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
690 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
692 opt_skb = skb_clone(skb, GFP_ATOMIC);
694 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
695 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
696 goto reset;
697 if (opt_skb) {
698 /* XXX This is where we would goto ipv6_pktoptions. */
699 __kfree_skb(opt_skb);
701 return 0;
705 * Step 3: Process LISTEN state
706 * If S.state == LISTEN,
707 * If P.type == Request or P contains a valid Init Cookie option,
708 * (* Must scan the packet's options to check for Init
709 * Cookies. Only Init Cookies are processed here,
710 * however; other options are processed in Step 8. This
711 * scan need only be performed if the endpoint uses Init
712 * Cookies *)
713 * (* Generate a new socket and switch to that socket *)
714 * Set S := new socket for this port pair
715 * S.state = RESPOND
716 * Choose S.ISS (initial seqno) or set from Init Cookies
717 * Initialize S.GAR := S.ISS
718 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
719 * Continue with S.state == RESPOND
720 * (* A Response packet will be generated in Step 11 *)
721 * Otherwise,
722 * Generate Reset(No Connection) unless P.type == Reset
723 * Drop packet and return
725 * NOTE: the check for the packet types is done in
726 * dccp_rcv_state_process
728 if (sk->sk_state == DCCP_LISTEN) {
729 struct sock *nsk = dccp_v6_hnd_req(sk, skb);
731 if (nsk == NULL)
732 goto discard;
734 * Queue it on the new socket if the new socket is active,
735 * otherwise we just shortcircuit this and continue with
736 * the new socket..
738 if (nsk != sk) {
739 if (dccp_child_process(sk, nsk, skb))
740 goto reset;
741 if (opt_skb != NULL)
742 __kfree_skb(opt_skb);
743 return 0;
747 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
748 goto reset;
749 if (opt_skb) {
750 /* XXX This is where we would goto ipv6_pktoptions. */
751 __kfree_skb(opt_skb);
753 return 0;
755 reset:
756 dccp_v6_ctl_send_reset(sk, skb);
757 discard:
758 if (opt_skb != NULL)
759 __kfree_skb(opt_skb);
760 kfree_skb(skb);
761 return 0;
764 static int dccp_v6_rcv(struct sk_buff *skb)
766 const struct dccp_hdr *dh;
767 struct sock *sk;
768 int min_cov;
770 /* Step 1: Check header basics */
772 if (dccp_invalid_packet(skb))
773 goto discard_it;
775 /* Step 1: If header checksum is incorrect, drop packet and return. */
776 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
777 &ipv6_hdr(skb)->daddr)) {
778 DCCP_WARN("dropped packet with invalid checksum\n");
779 goto discard_it;
782 dh = dccp_hdr(skb);
784 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
785 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
787 if (dccp_packet_without_ack(skb))
788 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
789 else
790 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
792 /* Step 2:
793 * Look up flow ID in table and get corresponding socket */
794 sk = __inet6_lookup(dev_net(skb->dst->dev), &dccp_hashinfo,
795 &ipv6_hdr(skb)->saddr, dh->dccph_sport,
796 &ipv6_hdr(skb)->daddr, ntohs(dh->dccph_dport),
797 inet6_iif(skb));
799 * Step 2:
800 * If no socket ...
802 if (sk == NULL) {
803 dccp_pr_debug("failed to look up flow ID in table and "
804 "get corresponding socket\n");
805 goto no_dccp_socket;
809 * Step 2:
810 * ... or S.state == TIMEWAIT,
811 * Generate Reset(No Connection) unless P.type == Reset
812 * Drop packet and return
814 if (sk->sk_state == DCCP_TIME_WAIT) {
815 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
816 inet_twsk_put(inet_twsk(sk));
817 goto no_dccp_socket;
821 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
822 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
823 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
825 min_cov = dccp_sk(sk)->dccps_pcrlen;
826 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
827 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
828 dh->dccph_cscov, min_cov);
829 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
830 goto discard_and_relse;
833 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
834 goto discard_and_relse;
836 return sk_receive_skb(sk, skb, 1) ? -1 : 0;
838 no_dccp_socket:
839 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
840 goto discard_it;
842 * Step 2:
843 * If no socket ...
844 * Generate Reset(No Connection) unless P.type == Reset
845 * Drop packet and return
847 if (dh->dccph_type != DCCP_PKT_RESET) {
848 DCCP_SKB_CB(skb)->dccpd_reset_code =
849 DCCP_RESET_CODE_NO_CONNECTION;
850 dccp_v6_ctl_send_reset(sk, skb);
853 discard_it:
854 kfree_skb(skb);
855 return 0;
857 discard_and_relse:
858 sock_put(sk);
859 goto discard_it;
862 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
863 int addr_len)
865 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
866 struct inet_connection_sock *icsk = inet_csk(sk);
867 struct inet_sock *inet = inet_sk(sk);
868 struct ipv6_pinfo *np = inet6_sk(sk);
869 struct dccp_sock *dp = dccp_sk(sk);
870 struct in6_addr *saddr = NULL, *final_p = NULL, final;
871 struct flowi fl;
872 struct dst_entry *dst;
873 int addr_type;
874 int err;
876 dp->dccps_role = DCCP_ROLE_CLIENT;
878 if (addr_len < SIN6_LEN_RFC2133)
879 return -EINVAL;
881 if (usin->sin6_family != AF_INET6)
882 return -EAFNOSUPPORT;
884 memset(&fl, 0, sizeof(fl));
886 if (np->sndflow) {
887 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
888 IP6_ECN_flow_init(fl.fl6_flowlabel);
889 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) {
890 struct ip6_flowlabel *flowlabel;
891 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
892 if (flowlabel == NULL)
893 return -EINVAL;
894 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
895 fl6_sock_release(flowlabel);
899 * connect() to INADDR_ANY means loopback (BSD'ism).
901 if (ipv6_addr_any(&usin->sin6_addr))
902 usin->sin6_addr.s6_addr[15] = 1;
904 addr_type = ipv6_addr_type(&usin->sin6_addr);
906 if (addr_type & IPV6_ADDR_MULTICAST)
907 return -ENETUNREACH;
909 if (addr_type & IPV6_ADDR_LINKLOCAL) {
910 if (addr_len >= sizeof(struct sockaddr_in6) &&
911 usin->sin6_scope_id) {
912 /* If interface is set while binding, indices
913 * must coincide.
915 if (sk->sk_bound_dev_if &&
916 sk->sk_bound_dev_if != usin->sin6_scope_id)
917 return -EINVAL;
919 sk->sk_bound_dev_if = usin->sin6_scope_id;
922 /* Connect to link-local address requires an interface */
923 if (!sk->sk_bound_dev_if)
924 return -EINVAL;
927 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
928 np->flow_label = fl.fl6_flowlabel;
931 * DCCP over IPv4
933 if (addr_type == IPV6_ADDR_MAPPED) {
934 u32 exthdrlen = icsk->icsk_ext_hdr_len;
935 struct sockaddr_in sin;
937 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
939 if (__ipv6_only_sock(sk))
940 return -ENETUNREACH;
942 sin.sin_family = AF_INET;
943 sin.sin_port = usin->sin6_port;
944 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
946 icsk->icsk_af_ops = &dccp_ipv6_mapped;
947 sk->sk_backlog_rcv = dccp_v4_do_rcv;
949 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
950 if (err) {
951 icsk->icsk_ext_hdr_len = exthdrlen;
952 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
953 sk->sk_backlog_rcv = dccp_v6_do_rcv;
954 goto failure;
955 } else {
956 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
957 inet->saddr);
958 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
959 inet->rcv_saddr);
962 return err;
965 if (!ipv6_addr_any(&np->rcv_saddr))
966 saddr = &np->rcv_saddr;
968 fl.proto = IPPROTO_DCCP;
969 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
970 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr);
971 fl.oif = sk->sk_bound_dev_if;
972 fl.fl_ip_dport = usin->sin6_port;
973 fl.fl_ip_sport = inet->sport;
974 security_sk_classify_flow(sk, &fl);
976 if (np->opt != NULL && np->opt->srcrt != NULL) {
977 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
979 ipv6_addr_copy(&final, &fl.fl6_dst);
980 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
981 final_p = &final;
984 err = ip6_dst_lookup(sk, &dst, &fl);
985 if (err)
986 goto failure;
988 if (final_p)
989 ipv6_addr_copy(&fl.fl6_dst, final_p);
991 err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT);
992 if (err < 0) {
993 if (err == -EREMOTE)
994 err = ip6_dst_blackhole(sk, &dst, &fl);
995 if (err < 0)
996 goto failure;
999 if (saddr == NULL) {
1000 saddr = &fl.fl6_src;
1001 ipv6_addr_copy(&np->rcv_saddr, saddr);
1004 /* set the source address */
1005 ipv6_addr_copy(&np->saddr, saddr);
1006 inet->rcv_saddr = LOOPBACK4_IPV6;
1008 __ip6_dst_store(sk, dst, NULL, NULL);
1010 icsk->icsk_ext_hdr_len = 0;
1011 if (np->opt != NULL)
1012 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
1013 np->opt->opt_nflen);
1015 inet->dport = usin->sin6_port;
1017 dccp_set_state(sk, DCCP_REQUESTING);
1018 err = inet6_hash_connect(&dccp_death_row, sk);
1019 if (err)
1020 goto late_failure;
1022 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
1023 np->daddr.s6_addr32,
1024 inet->sport, inet->dport);
1025 err = dccp_connect(sk);
1026 if (err)
1027 goto late_failure;
1029 return 0;
1031 late_failure:
1032 dccp_set_state(sk, DCCP_CLOSED);
1033 __sk_dst_reset(sk);
1034 failure:
1035 inet->dport = 0;
1036 sk->sk_route_caps = 0;
1037 return err;
1040 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
1041 .queue_xmit = inet6_csk_xmit,
1042 .send_check = dccp_v6_send_check,
1043 .rebuild_header = inet6_sk_rebuild_header,
1044 .conn_request = dccp_v6_conn_request,
1045 .syn_recv_sock = dccp_v6_request_recv_sock,
1046 .net_header_len = sizeof(struct ipv6hdr),
1047 .setsockopt = ipv6_setsockopt,
1048 .getsockopt = ipv6_getsockopt,
1049 .addr2sockaddr = inet6_csk_addr2sockaddr,
1050 .sockaddr_len = sizeof(struct sockaddr_in6),
1051 .bind_conflict = inet6_csk_bind_conflict,
1052 #ifdef CONFIG_COMPAT
1053 .compat_setsockopt = compat_ipv6_setsockopt,
1054 .compat_getsockopt = compat_ipv6_getsockopt,
1055 #endif
1059 * DCCP over IPv4 via INET6 API
1061 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
1062 .queue_xmit = ip_queue_xmit,
1063 .send_check = dccp_v4_send_check,
1064 .rebuild_header = inet_sk_rebuild_header,
1065 .conn_request = dccp_v6_conn_request,
1066 .syn_recv_sock = dccp_v6_request_recv_sock,
1067 .net_header_len = sizeof(struct iphdr),
1068 .setsockopt = ipv6_setsockopt,
1069 .getsockopt = ipv6_getsockopt,
1070 .addr2sockaddr = inet6_csk_addr2sockaddr,
1071 .sockaddr_len = sizeof(struct sockaddr_in6),
1072 #ifdef CONFIG_COMPAT
1073 .compat_setsockopt = compat_ipv6_setsockopt,
1074 .compat_getsockopt = compat_ipv6_getsockopt,
1075 #endif
1078 /* NOTE: A lot of things set to zero explicitly by call to
1079 * sk_alloc() so need not be done here.
1081 static int dccp_v6_init_sock(struct sock *sk)
1083 static __u8 dccp_v6_ctl_sock_initialized;
1084 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1086 if (err == 0) {
1087 if (unlikely(!dccp_v6_ctl_sock_initialized))
1088 dccp_v6_ctl_sock_initialized = 1;
1089 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1092 return err;
1095 static int dccp_v6_destroy_sock(struct sock *sk)
1097 dccp_destroy_sock(sk);
1098 return inet6_destroy_sock(sk);
1101 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1102 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1105 static struct proto dccp_v6_prot = {
1106 .name = "DCCPv6",
1107 .owner = THIS_MODULE,
1108 .close = dccp_close,
1109 .connect = dccp_v6_connect,
1110 .disconnect = dccp_disconnect,
1111 .ioctl = dccp_ioctl,
1112 .init = dccp_v6_init_sock,
1113 .setsockopt = dccp_setsockopt,
1114 .getsockopt = dccp_getsockopt,
1115 .sendmsg = dccp_sendmsg,
1116 .recvmsg = dccp_recvmsg,
1117 .backlog_rcv = dccp_v6_do_rcv,
1118 .hash = dccp_v6_hash,
1119 .unhash = inet_unhash,
1120 .accept = inet_csk_accept,
1121 .get_port = inet_csk_get_port,
1122 .shutdown = dccp_shutdown,
1123 .destroy = dccp_v6_destroy_sock,
1124 .orphan_count = &dccp_orphan_count,
1125 .max_header = MAX_DCCP_HEADER,
1126 .obj_size = sizeof(struct dccp6_sock),
1127 .rsk_prot = &dccp6_request_sock_ops,
1128 .twsk_prot = &dccp6_timewait_sock_ops,
1129 .h.hashinfo = &dccp_hashinfo,
1130 #ifdef CONFIG_COMPAT
1131 .compat_setsockopt = compat_dccp_setsockopt,
1132 .compat_getsockopt = compat_dccp_getsockopt,
1133 #endif
1136 static struct inet6_protocol dccp_v6_protocol = {
1137 .handler = dccp_v6_rcv,
1138 .err_handler = dccp_v6_err,
1139 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1142 static struct proto_ops inet6_dccp_ops = {
1143 .family = PF_INET6,
1144 .owner = THIS_MODULE,
1145 .release = inet6_release,
1146 .bind = inet6_bind,
1147 .connect = inet_stream_connect,
1148 .socketpair = sock_no_socketpair,
1149 .accept = inet_accept,
1150 .getname = inet6_getname,
1151 .poll = dccp_poll,
1152 .ioctl = inet6_ioctl,
1153 .listen = inet_dccp_listen,
1154 .shutdown = inet_shutdown,
1155 .setsockopt = sock_common_setsockopt,
1156 .getsockopt = sock_common_getsockopt,
1157 .sendmsg = inet_sendmsg,
1158 .recvmsg = sock_common_recvmsg,
1159 .mmap = sock_no_mmap,
1160 .sendpage = sock_no_sendpage,
1161 #ifdef CONFIG_COMPAT
1162 .compat_setsockopt = compat_sock_common_setsockopt,
1163 .compat_getsockopt = compat_sock_common_getsockopt,
1164 #endif
1167 static struct inet_protosw dccp_v6_protosw = {
1168 .type = SOCK_DCCP,
1169 .protocol = IPPROTO_DCCP,
1170 .prot = &dccp_v6_prot,
1171 .ops = &inet6_dccp_ops,
1172 .capability = -1,
1173 .flags = INET_PROTOSW_ICSK,
1176 static int dccp_v6_init_net(struct net *net)
1178 int err;
1180 err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1181 SOCK_DCCP, IPPROTO_DCCP, net);
1182 return err;
1185 static void dccp_v6_exit_net(struct net *net)
1187 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1190 static struct pernet_operations dccp_v6_ops = {
1191 .init = dccp_v6_init_net,
1192 .exit = dccp_v6_exit_net,
1195 static int __init dccp_v6_init(void)
1197 int err = proto_register(&dccp_v6_prot, 1);
1199 if (err != 0)
1200 goto out;
1202 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1203 if (err != 0)
1204 goto out_unregister_proto;
1206 inet6_register_protosw(&dccp_v6_protosw);
1208 err = register_pernet_subsys(&dccp_v6_ops);
1209 if (err != 0)
1210 goto out_destroy_ctl_sock;
1211 out:
1212 return err;
1214 out_destroy_ctl_sock:
1215 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1216 inet6_unregister_protosw(&dccp_v6_protosw);
1217 out_unregister_proto:
1218 proto_unregister(&dccp_v6_prot);
1219 goto out;
1222 static void __exit dccp_v6_exit(void)
1224 unregister_pernet_subsys(&dccp_v6_ops);
1225 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1226 inet6_unregister_protosw(&dccp_v6_protosw);
1227 proto_unregister(&dccp_v6_prot);
1230 module_init(dccp_v6_init);
1231 module_exit(dccp_v6_exit);
1234 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1235 * values directly, Also cover the case where the protocol is not specified,
1236 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1238 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1239 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1240 MODULE_LICENSE("GPL");
1241 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1242 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");