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>
27 #include <net/protocol.h>
28 #include <net/transp_v6.h>
29 #include <net/ip6_checksum.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
) {
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
,
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
);
93 struct ipv6_pinfo
*np
;
97 struct net
*net
= dev_net(skb
->dev
);
99 sk
= inet6_lookup(net
, &dccp_hashinfo
,
100 &hdr
->daddr
, dh
->dccph_dport
,
101 &hdr
->saddr
, dh
->dccph_sport
, inet6_iif(skb
));
104 ICMP6_INC_STATS_BH(__in6_dev_get(skb
->dev
), ICMP6_MIB_INERRORS
);
108 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
109 inet_twsk_put(inet_twsk(sk
));
114 if (sock_owned_by_user(sk
))
115 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
117 if (sk
->sk_state
== DCCP_CLOSED
)
121 seq
= dccp_hdr_seq(dh
);
122 if ((1 << sk
->sk_state
) & ~(DCCPF_REQUESTING
| DCCPF_LISTEN
) &&
123 !between48(seq
, dp
->dccps_awl
, dp
->dccps_awh
)) {
124 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
130 if (type
== ICMPV6_PKT_TOOBIG
) {
131 struct dst_entry
*dst
= NULL
;
133 if (sock_owned_by_user(sk
))
135 if ((1 << sk
->sk_state
) & (DCCPF_LISTEN
| DCCPF_CLOSED
))
138 /* icmp should have updated the destination cache entry */
139 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
141 struct inet_sock
*inet
= inet_sk(sk
);
144 /* BUGGG_FUTURE: Again, it is not clear how
145 to handle rthdr case. Ignore this complexity
148 memset(&fl
, 0, sizeof(fl
));
149 fl
.proto
= IPPROTO_DCCP
;
150 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
151 ipv6_addr_copy(&fl
.fl6_src
, &np
->saddr
);
152 fl
.oif
= sk
->sk_bound_dev_if
;
153 fl
.fl_ip_dport
= inet
->dport
;
154 fl
.fl_ip_sport
= inet
->sport
;
155 security_sk_classify_flow(sk
, &fl
);
157 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
159 sk
->sk_err_soft
= -err
;
163 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
165 sk
->sk_err_soft
= -err
;
171 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
)) {
172 dccp_sync_mss(sk
, dst_mtu(dst
));
173 } /* else let the usual retransmit timer handle it */
178 icmpv6_err_convert(type
, code
, &err
);
180 /* Might be for an request_sock */
181 switch (sk
->sk_state
) {
182 struct request_sock
*req
, **prev
;
184 if (sock_owned_by_user(sk
))
187 req
= inet6_csk_search_req(sk
, &prev
, dh
->dccph_dport
,
188 &hdr
->daddr
, &hdr
->saddr
,
194 * ICMPs are not backlogged, hence we cannot get an established
197 WARN_ON(req
->sk
!= NULL
);
199 if (seq
!= dccp_rsk(req
)->dreq_iss
) {
200 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
204 inet_csk_reqsk_queue_drop(sk
, req
, prev
);
207 case DCCP_REQUESTING
:
208 case DCCP_RESPOND
: /* Cannot happen.
209 It can, it SYNs are crossed. --ANK */
210 if (!sock_owned_by_user(sk
)) {
211 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
214 * Wake people up to see the error
215 * (see connect in sock.c)
217 sk
->sk_error_report(sk
);
220 sk
->sk_err_soft
= err
;
224 if (!sock_owned_by_user(sk
) && np
->recverr
) {
226 sk
->sk_error_report(sk
);
228 sk
->sk_err_soft
= err
;
236 static int dccp_v6_send_response(struct sock
*sk
, struct request_sock
*req
)
238 struct inet6_request_sock
*ireq6
= inet6_rsk(req
);
239 struct ipv6_pinfo
*np
= inet6_sk(sk
);
241 struct ipv6_txoptions
*opt
= NULL
;
242 struct in6_addr
*final_p
= NULL
, final
;
245 struct dst_entry
*dst
;
247 memset(&fl
, 0, sizeof(fl
));
248 fl
.proto
= IPPROTO_DCCP
;
249 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
250 ipv6_addr_copy(&fl
.fl6_src
, &ireq6
->loc_addr
);
251 fl
.fl6_flowlabel
= 0;
253 fl
.fl_ip_dport
= inet_rsk(req
)->rmt_port
;
254 fl
.fl_ip_sport
= inet_sk(sk
)->sport
;
255 security_req_classify_flow(req
, &fl
);
259 if (opt
!= NULL
&& opt
->srcrt
!= NULL
) {
260 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)opt
->srcrt
;
262 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
263 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
267 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
272 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
274 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
278 skb
= dccp_make_response(sk
, dst
, req
);
280 struct dccp_hdr
*dh
= dccp_hdr(skb
);
282 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
,
285 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
286 err
= ip6_xmit(sk
, skb
, &fl
, opt
, 0);
287 err
= net_xmit_eval(err
);
291 if (opt
!= NULL
&& opt
!= np
->opt
)
292 sock_kfree_s(sk
, opt
, opt
->tot_len
);
297 static void dccp_v6_reqsk_destructor(struct request_sock
*req
)
299 if (inet6_rsk(req
)->pktopts
!= NULL
)
300 kfree_skb(inet6_rsk(req
)->pktopts
);
303 static void dccp_v6_ctl_send_reset(struct sock
*sk
, struct sk_buff
*rxskb
)
305 struct ipv6hdr
*rxip6h
;
308 struct net
*net
= dev_net(rxskb
->dst
->dev
);
309 struct sock
*ctl_sk
= net
->dccp
.v6_ctl_sk
;
311 if (dccp_hdr(rxskb
)->dccph_type
== DCCP_PKT_RESET
)
314 if (!ipv6_unicast_destination(rxskb
))
317 skb
= dccp_ctl_make_reset(ctl_sk
, rxskb
);
321 rxip6h
= ipv6_hdr(rxskb
);
322 dccp_hdr(skb
)->dccph_checksum
= dccp_v6_csum_finish(skb
, &rxip6h
->saddr
,
325 memset(&fl
, 0, sizeof(fl
));
326 ipv6_addr_copy(&fl
.fl6_dst
, &rxip6h
->saddr
);
327 ipv6_addr_copy(&fl
.fl6_src
, &rxip6h
->daddr
);
329 fl
.proto
= IPPROTO_DCCP
;
330 fl
.oif
= inet6_iif(rxskb
);
331 fl
.fl_ip_dport
= dccp_hdr(skb
)->dccph_dport
;
332 fl
.fl_ip_sport
= dccp_hdr(skb
)->dccph_sport
;
333 security_skb_classify_flow(rxskb
, &fl
);
335 /* sk = NULL, but it is safe for now. RST socket required. */
336 if (!ip6_dst_lookup(ctl_sk
, &skb
->dst
, &fl
)) {
337 if (xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) >= 0) {
338 ip6_xmit(ctl_sk
, skb
, &fl
, NULL
, 0);
339 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS
);
340 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS
);
348 static struct request_sock_ops dccp6_request_sock_ops
= {
350 .obj_size
= sizeof(struct dccp6_request_sock
),
351 .rtx_syn_ack
= dccp_v6_send_response
,
352 .send_ack
= dccp_reqsk_send_ack
,
353 .destructor
= dccp_v6_reqsk_destructor
,
354 .send_reset
= dccp_v6_ctl_send_reset
,
357 static struct sock
*dccp_v6_hnd_req(struct sock
*sk
,struct sk_buff
*skb
)
359 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
360 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
362 struct request_sock
**prev
;
363 /* Find possible connection requests. */
364 struct request_sock
*req
= inet6_csk_search_req(sk
, &prev
,
370 return dccp_check_req(sk
, skb
, req
, prev
);
372 nsk
= __inet6_lookup_established(sock_net(sk
), &dccp_hashinfo
,
373 &iph
->saddr
, dh
->dccph_sport
,
374 &iph
->daddr
, ntohs(dh
->dccph_dport
),
377 if (nsk
->sk_state
!= DCCP_TIME_WAIT
) {
381 inet_twsk_put(inet_twsk(nsk
));
388 static int dccp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
390 struct request_sock
*req
;
391 struct dccp_request_sock
*dreq
;
392 struct inet6_request_sock
*ireq6
;
393 struct ipv6_pinfo
*np
= inet6_sk(sk
);
394 const __be32 service
= dccp_hdr_request(skb
)->dccph_req_service
;
395 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
397 if (skb
->protocol
== htons(ETH_P_IP
))
398 return dccp_v4_conn_request(sk
, skb
);
400 if (!ipv6_unicast_destination(skb
))
401 return 0; /* discard, don't send a reset here */
403 if (dccp_bad_service_code(sk
, service
)) {
404 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_BAD_SERVICE_CODE
;
408 * There are no SYN attacks on IPv6, yet...
410 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_TOO_BUSY
;
411 if (inet_csk_reqsk_queue_is_full(sk
))
414 if (sk_acceptq_is_full(sk
) && inet_csk_reqsk_queue_young(sk
) > 1)
417 req
= inet6_reqsk_alloc(&dccp6_request_sock_ops
);
421 dccp_reqsk_init(req
, skb
);
423 dreq
= dccp_rsk(req
);
424 if (dccp_parse_options(sk
, dreq
, skb
))
427 if (security_inet_conn_request(sk
, skb
, req
))
430 ireq6
= inet6_rsk(req
);
431 ipv6_addr_copy(&ireq6
->rmt_addr
, &ipv6_hdr(skb
)->saddr
);
432 ipv6_addr_copy(&ireq6
->loc_addr
, &ipv6_hdr(skb
)->daddr
);
434 if (ipv6_opt_accepted(sk
, skb
) ||
435 np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
||
436 np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
) {
437 atomic_inc(&skb
->users
);
438 ireq6
->pktopts
= skb
;
440 ireq6
->iif
= sk
->sk_bound_dev_if
;
442 /* So that link locals have meaning */
443 if (!sk
->sk_bound_dev_if
&&
444 ipv6_addr_type(&ireq6
->rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
445 ireq6
->iif
= inet6_iif(skb
);
448 * Step 3: Process LISTEN state
450 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
452 * In fact we defer setting S.GSR, S.SWL, S.SWH to
453 * dccp_create_openreq_child.
455 dreq
->dreq_isr
= dcb
->dccpd_seq
;
456 dreq
->dreq_iss
= dccp_v6_init_sequence(skb
);
457 dreq
->dreq_service
= service
;
459 if (dccp_v6_send_response(sk
, req
))
462 inet6_csk_reqsk_queue_hash_add(sk
, req
, DCCP_TIMEOUT_INIT
);
468 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
472 static struct sock
*dccp_v6_request_recv_sock(struct sock
*sk
,
474 struct request_sock
*req
,
475 struct dst_entry
*dst
)
477 struct inet6_request_sock
*ireq6
= inet6_rsk(req
);
478 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
479 struct inet_sock
*newinet
;
480 struct dccp_sock
*newdp
;
481 struct dccp6_sock
*newdp6
;
483 struct ipv6_txoptions
*opt
;
485 if (skb
->protocol
== htons(ETH_P_IP
)) {
489 newsk
= dccp_v4_request_recv_sock(sk
, skb
, req
, dst
);
493 newdp6
= (struct dccp6_sock
*)newsk
;
494 newdp
= dccp_sk(newsk
);
495 newinet
= inet_sk(newsk
);
496 newinet
->pinet6
= &newdp6
->inet6
;
497 newnp
= inet6_sk(newsk
);
499 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
501 ipv6_addr_set(&newnp
->daddr
, 0, 0, htonl(0x0000FFFF),
504 ipv6_addr_set(&newnp
->saddr
, 0, 0, htonl(0x0000FFFF),
507 ipv6_addr_copy(&newnp
->rcv_saddr
, &newnp
->saddr
);
509 inet_csk(newsk
)->icsk_af_ops
= &dccp_ipv6_mapped
;
510 newsk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
511 newnp
->pktoptions
= NULL
;
513 newnp
->mcast_oif
= inet6_iif(skb
);
514 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
517 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
518 * here, dccp_create_openreq_child now does this for us, see the comment in
519 * that function for the gory details. -acme
522 /* It is tricky place. Until this moment IPv4 tcp
523 worked with IPv6 icsk.icsk_af_ops.
526 dccp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
533 if (sk_acceptq_is_full(sk
))
537 struct in6_addr
*final_p
= NULL
, final
;
540 memset(&fl
, 0, sizeof(fl
));
541 fl
.proto
= IPPROTO_DCCP
;
542 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
543 if (opt
!= NULL
&& opt
->srcrt
!= NULL
) {
544 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)opt
->srcrt
;
546 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
547 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
550 ipv6_addr_copy(&fl
.fl6_src
, &ireq6
->loc_addr
);
551 fl
.oif
= sk
->sk_bound_dev_if
;
552 fl
.fl_ip_dport
= inet_rsk(req
)->rmt_port
;
553 fl
.fl_ip_sport
= inet_sk(sk
)->sport
;
554 security_sk_classify_flow(sk
, &fl
);
556 if (ip6_dst_lookup(sk
, &dst
, &fl
))
560 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
562 if ((xfrm_lookup(&dst
, &fl
, sk
, 0)) < 0)
566 newsk
= dccp_create_openreq_child(sk
, req
, skb
);
571 * No need to charge this sock to the relevant IPv6 refcnt debug socks
572 * count here, dccp_create_openreq_child now does this for us, see the
573 * comment in that function for the gory details. -acme
576 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
577 newsk
->sk_route_caps
= dst
->dev
->features
& ~(NETIF_F_IP_CSUM
|
579 newdp6
= (struct dccp6_sock
*)newsk
;
580 newinet
= inet_sk(newsk
);
581 newinet
->pinet6
= &newdp6
->inet6
;
582 newdp
= dccp_sk(newsk
);
583 newnp
= inet6_sk(newsk
);
585 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
587 ipv6_addr_copy(&newnp
->daddr
, &ireq6
->rmt_addr
);
588 ipv6_addr_copy(&newnp
->saddr
, &ireq6
->loc_addr
);
589 ipv6_addr_copy(&newnp
->rcv_saddr
, &ireq6
->loc_addr
);
590 newsk
->sk_bound_dev_if
= ireq6
->iif
;
592 /* Now IPv6 options...
594 First: no IPv4 options.
599 newnp
->rxopt
.all
= np
->rxopt
.all
;
601 /* Clone pktoptions received with SYN */
602 newnp
->pktoptions
= NULL
;
603 if (ireq6
->pktopts
!= NULL
) {
604 newnp
->pktoptions
= skb_clone(ireq6
->pktopts
, GFP_ATOMIC
);
605 kfree_skb(ireq6
->pktopts
);
606 ireq6
->pktopts
= NULL
;
607 if (newnp
->pktoptions
)
608 skb_set_owner_r(newnp
->pktoptions
, newsk
);
611 newnp
->mcast_oif
= inet6_iif(skb
);
612 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
615 * Clone native IPv6 options from listening socket (if any)
617 * Yes, keeping reference count would be much more clever, but we make
618 * one more one thing there: reattach optmem to newsk.
621 newnp
->opt
= ipv6_dup_options(newsk
, opt
);
623 sock_kfree_s(sk
, opt
, opt
->tot_len
);
626 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
627 if (newnp
->opt
!= NULL
)
628 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
629 newnp
->opt
->opt_flen
);
631 dccp_sync_mss(newsk
, dst_mtu(dst
));
633 newinet
->daddr
= newinet
->saddr
= newinet
->rcv_saddr
= LOOPBACK4_IPV6
;
636 __inet_inherit_port(sk
, newsk
);
641 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
643 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
644 if (opt
!= NULL
&& opt
!= np
->opt
)
645 sock_kfree_s(sk
, opt
, opt
->tot_len
);
650 /* The socket must have it's spinlock held when we get
653 * We have a potential double-lock case here, so even when
654 * doing backlog processing we use the BH locking scheme.
655 * This is because we cannot sleep with the original spinlock
658 static int dccp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
660 struct ipv6_pinfo
*np
= inet6_sk(sk
);
661 struct sk_buff
*opt_skb
= NULL
;
663 /* Imagine: socket is IPv6. IPv4 packet arrives,
664 goes to IPv4 receive handler and backlogged.
665 From backlog it always goes here. Kerboom...
666 Fortunately, dccp_rcv_established and rcv_established
667 handle them correctly, but it is not case with
668 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
671 if (skb
->protocol
== htons(ETH_P_IP
))
672 return dccp_v4_do_rcv(sk
, skb
);
674 if (sk_filter(sk
, skb
))
678 * socket locking is here for SMP purposes as backlog rcv is currently
679 * called with bh processing disabled.
682 /* Do Stevens' IPV6_PKTOPTIONS.
684 Yes, guys, it is the only place in our code, where we
685 may make it not affecting IPv4.
686 The rest of code is protocol independent,
687 and I do not like idea to uglify IPv4.
689 Actually, all the idea behind IPV6_PKTOPTIONS
690 looks not very well thought. For now we latch
691 options, received in the last packet, enqueued
692 by tcp. Feel free to propose better solution.
697 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
698 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
700 opt_skb
= skb_clone(skb
, GFP_ATOMIC
);
702 if (sk
->sk_state
== DCCP_OPEN
) { /* Fast path */
703 if (dccp_rcv_established(sk
, skb
, dccp_hdr(skb
), skb
->len
))
706 /* XXX This is where we would goto ipv6_pktoptions. */
707 __kfree_skb(opt_skb
);
713 * Step 3: Process LISTEN state
714 * If S.state == LISTEN,
715 * If P.type == Request or P contains a valid Init Cookie option,
716 * (* Must scan the packet's options to check for Init
717 * Cookies. Only Init Cookies are processed here,
718 * however; other options are processed in Step 8. This
719 * scan need only be performed if the endpoint uses Init
721 * (* Generate a new socket and switch to that socket *)
722 * Set S := new socket for this port pair
724 * Choose S.ISS (initial seqno) or set from Init Cookies
725 * Initialize S.GAR := S.ISS
726 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
727 * Continue with S.state == RESPOND
728 * (* A Response packet will be generated in Step 11 *)
730 * Generate Reset(No Connection) unless P.type == Reset
731 * Drop packet and return
733 * NOTE: the check for the packet types is done in
734 * dccp_rcv_state_process
736 if (sk
->sk_state
== DCCP_LISTEN
) {
737 struct sock
*nsk
= dccp_v6_hnd_req(sk
, skb
);
742 * Queue it on the new socket if the new socket is active,
743 * otherwise we just shortcircuit this and continue with
747 if (dccp_child_process(sk
, nsk
, skb
))
750 __kfree_skb(opt_skb
);
755 if (dccp_rcv_state_process(sk
, skb
, dccp_hdr(skb
), skb
->len
))
758 /* XXX This is where we would goto ipv6_pktoptions. */
759 __kfree_skb(opt_skb
);
764 dccp_v6_ctl_send_reset(sk
, skb
);
767 __kfree_skb(opt_skb
);
772 static int dccp_v6_rcv(struct sk_buff
*skb
)
774 const struct dccp_hdr
*dh
;
778 /* Step 1: Check header basics */
780 if (dccp_invalid_packet(skb
))
783 /* Step 1: If header checksum is incorrect, drop packet and return. */
784 if (dccp_v6_csum_finish(skb
, &ipv6_hdr(skb
)->saddr
,
785 &ipv6_hdr(skb
)->daddr
)) {
786 DCCP_WARN("dropped packet with invalid checksum\n");
792 DCCP_SKB_CB(skb
)->dccpd_seq
= dccp_hdr_seq(dh
);
793 DCCP_SKB_CB(skb
)->dccpd_type
= dh
->dccph_type
;
795 if (dccp_packet_without_ack(skb
))
796 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= DCCP_PKT_WITHOUT_ACK_SEQ
;
798 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= dccp_hdr_ack_seq(skb
);
801 * Look up flow ID in table and get corresponding socket */
802 sk
= __inet6_lookup(dev_net(skb
->dst
->dev
), &dccp_hashinfo
,
803 &ipv6_hdr(skb
)->saddr
, dh
->dccph_sport
,
804 &ipv6_hdr(skb
)->daddr
, ntohs(dh
->dccph_dport
),
811 dccp_pr_debug("failed to look up flow ID in table and "
812 "get corresponding socket\n");
818 * ... or S.state == TIMEWAIT,
819 * Generate Reset(No Connection) unless P.type == Reset
820 * Drop packet and return
822 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
823 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
824 inet_twsk_put(inet_twsk(sk
));
829 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
830 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
831 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
833 min_cov
= dccp_sk(sk
)->dccps_pcrlen
;
834 if (dh
->dccph_cscov
&& (min_cov
== 0 || dh
->dccph_cscov
< min_cov
)) {
835 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
836 dh
->dccph_cscov
, min_cov
);
837 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
838 goto discard_and_relse
;
841 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
842 goto discard_and_relse
;
844 return sk_receive_skb(sk
, skb
, 1) ? -1 : 0;
847 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
852 * Generate Reset(No Connection) unless P.type == Reset
853 * Drop packet and return
855 if (dh
->dccph_type
!= DCCP_PKT_RESET
) {
856 DCCP_SKB_CB(skb
)->dccpd_reset_code
=
857 DCCP_RESET_CODE_NO_CONNECTION
;
858 dccp_v6_ctl_send_reset(sk
, skb
);
870 static int dccp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
873 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*)uaddr
;
874 struct inet_connection_sock
*icsk
= inet_csk(sk
);
875 struct inet_sock
*inet
= inet_sk(sk
);
876 struct ipv6_pinfo
*np
= inet6_sk(sk
);
877 struct dccp_sock
*dp
= dccp_sk(sk
);
878 struct in6_addr
*saddr
= NULL
, *final_p
= NULL
, final
;
880 struct dst_entry
*dst
;
884 dp
->dccps_role
= DCCP_ROLE_CLIENT
;
886 if (addr_len
< SIN6_LEN_RFC2133
)
889 if (usin
->sin6_family
!= AF_INET6
)
890 return -EAFNOSUPPORT
;
892 memset(&fl
, 0, sizeof(fl
));
895 fl
.fl6_flowlabel
= usin
->sin6_flowinfo
& IPV6_FLOWINFO_MASK
;
896 IP6_ECN_flow_init(fl
.fl6_flowlabel
);
897 if (fl
.fl6_flowlabel
& IPV6_FLOWLABEL_MASK
) {
898 struct ip6_flowlabel
*flowlabel
;
899 flowlabel
= fl6_sock_lookup(sk
, fl
.fl6_flowlabel
);
900 if (flowlabel
== NULL
)
902 ipv6_addr_copy(&usin
->sin6_addr
, &flowlabel
->dst
);
903 fl6_sock_release(flowlabel
);
907 * connect() to INADDR_ANY means loopback (BSD'ism).
909 if (ipv6_addr_any(&usin
->sin6_addr
))
910 usin
->sin6_addr
.s6_addr
[15] = 1;
912 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
914 if (addr_type
& IPV6_ADDR_MULTICAST
)
917 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
918 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
919 usin
->sin6_scope_id
) {
920 /* If interface is set while binding, indices
923 if (sk
->sk_bound_dev_if
&&
924 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
927 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
930 /* Connect to link-local address requires an interface */
931 if (!sk
->sk_bound_dev_if
)
935 ipv6_addr_copy(&np
->daddr
, &usin
->sin6_addr
);
936 np
->flow_label
= fl
.fl6_flowlabel
;
941 if (addr_type
== IPV6_ADDR_MAPPED
) {
942 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
943 struct sockaddr_in sin
;
945 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
947 if (__ipv6_only_sock(sk
))
950 sin
.sin_family
= AF_INET
;
951 sin
.sin_port
= usin
->sin6_port
;
952 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
954 icsk
->icsk_af_ops
= &dccp_ipv6_mapped
;
955 sk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
957 err
= dccp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
959 icsk
->icsk_ext_hdr_len
= exthdrlen
;
960 icsk
->icsk_af_ops
= &dccp_ipv6_af_ops
;
961 sk
->sk_backlog_rcv
= dccp_v6_do_rcv
;
964 ipv6_addr_set(&np
->saddr
, 0, 0, htonl(0x0000FFFF),
966 ipv6_addr_set(&np
->rcv_saddr
, 0, 0, htonl(0x0000FFFF),
973 if (!ipv6_addr_any(&np
->rcv_saddr
))
974 saddr
= &np
->rcv_saddr
;
976 fl
.proto
= IPPROTO_DCCP
;
977 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
978 ipv6_addr_copy(&fl
.fl6_src
, saddr
? saddr
: &np
->saddr
);
979 fl
.oif
= sk
->sk_bound_dev_if
;
980 fl
.fl_ip_dport
= usin
->sin6_port
;
981 fl
.fl_ip_sport
= inet
->sport
;
982 security_sk_classify_flow(sk
, &fl
);
984 if (np
->opt
!= NULL
&& np
->opt
->srcrt
!= NULL
) {
985 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)np
->opt
->srcrt
;
987 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
988 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
992 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
997 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
999 err
= __xfrm_lookup(&dst
, &fl
, sk
, XFRM_LOOKUP_WAIT
);
1001 if (err
== -EREMOTE
)
1002 err
= ip6_dst_blackhole(sk
, &dst
, &fl
);
1007 if (saddr
== NULL
) {
1008 saddr
= &fl
.fl6_src
;
1009 ipv6_addr_copy(&np
->rcv_saddr
, saddr
);
1012 /* set the source address */
1013 ipv6_addr_copy(&np
->saddr
, saddr
);
1014 inet
->rcv_saddr
= LOOPBACK4_IPV6
;
1016 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
1018 icsk
->icsk_ext_hdr_len
= 0;
1019 if (np
->opt
!= NULL
)
1020 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
1021 np
->opt
->opt_nflen
);
1023 inet
->dport
= usin
->sin6_port
;
1025 dccp_set_state(sk
, DCCP_REQUESTING
);
1026 err
= inet6_hash_connect(&dccp_death_row
, sk
);
1030 dp
->dccps_iss
= secure_dccpv6_sequence_number(np
->saddr
.s6_addr32
,
1031 np
->daddr
.s6_addr32
,
1032 inet
->sport
, inet
->dport
);
1033 err
= dccp_connect(sk
);
1040 dccp_set_state(sk
, DCCP_CLOSED
);
1044 sk
->sk_route_caps
= 0;
1048 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops
= {
1049 .queue_xmit
= inet6_csk_xmit
,
1050 .send_check
= dccp_v6_send_check
,
1051 .rebuild_header
= inet6_sk_rebuild_header
,
1052 .conn_request
= dccp_v6_conn_request
,
1053 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1054 .net_header_len
= sizeof(struct ipv6hdr
),
1055 .setsockopt
= ipv6_setsockopt
,
1056 .getsockopt
= ipv6_getsockopt
,
1057 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1058 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1059 .bind_conflict
= inet6_csk_bind_conflict
,
1060 #ifdef CONFIG_COMPAT
1061 .compat_setsockopt
= compat_ipv6_setsockopt
,
1062 .compat_getsockopt
= compat_ipv6_getsockopt
,
1067 * DCCP over IPv4 via INET6 API
1069 static struct inet_connection_sock_af_ops dccp_ipv6_mapped
= {
1070 .queue_xmit
= ip_queue_xmit
,
1071 .send_check
= dccp_v4_send_check
,
1072 .rebuild_header
= inet_sk_rebuild_header
,
1073 .conn_request
= dccp_v6_conn_request
,
1074 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1075 .net_header_len
= sizeof(struct iphdr
),
1076 .setsockopt
= ipv6_setsockopt
,
1077 .getsockopt
= ipv6_getsockopt
,
1078 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1079 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1080 #ifdef CONFIG_COMPAT
1081 .compat_setsockopt
= compat_ipv6_setsockopt
,
1082 .compat_getsockopt
= compat_ipv6_getsockopt
,
1086 /* NOTE: A lot of things set to zero explicitly by call to
1087 * sk_alloc() so need not be done here.
1089 static int dccp_v6_init_sock(struct sock
*sk
)
1091 static __u8 dccp_v6_ctl_sock_initialized
;
1092 int err
= dccp_init_sock(sk
, dccp_v6_ctl_sock_initialized
);
1095 if (unlikely(!dccp_v6_ctl_sock_initialized
))
1096 dccp_v6_ctl_sock_initialized
= 1;
1097 inet_csk(sk
)->icsk_af_ops
= &dccp_ipv6_af_ops
;
1103 static void dccp_v6_destroy_sock(struct sock
*sk
)
1105 dccp_destroy_sock(sk
);
1106 inet6_destroy_sock(sk
);
1109 static struct timewait_sock_ops dccp6_timewait_sock_ops
= {
1110 .twsk_obj_size
= sizeof(struct dccp6_timewait_sock
),
1113 static struct proto dccp_v6_prot
= {
1115 .owner
= THIS_MODULE
,
1116 .close
= dccp_close
,
1117 .connect
= dccp_v6_connect
,
1118 .disconnect
= dccp_disconnect
,
1119 .ioctl
= dccp_ioctl
,
1120 .init
= dccp_v6_init_sock
,
1121 .setsockopt
= dccp_setsockopt
,
1122 .getsockopt
= dccp_getsockopt
,
1123 .sendmsg
= dccp_sendmsg
,
1124 .recvmsg
= dccp_recvmsg
,
1125 .backlog_rcv
= dccp_v6_do_rcv
,
1126 .hash
= dccp_v6_hash
,
1127 .unhash
= inet_unhash
,
1128 .accept
= inet_csk_accept
,
1129 .get_port
= inet_csk_get_port
,
1130 .shutdown
= dccp_shutdown
,
1131 .destroy
= dccp_v6_destroy_sock
,
1132 .orphan_count
= &dccp_orphan_count
,
1133 .max_header
= MAX_DCCP_HEADER
,
1134 .obj_size
= sizeof(struct dccp6_sock
),
1135 .rsk_prot
= &dccp6_request_sock_ops
,
1136 .twsk_prot
= &dccp6_timewait_sock_ops
,
1137 .h
.hashinfo
= &dccp_hashinfo
,
1138 #ifdef CONFIG_COMPAT
1139 .compat_setsockopt
= compat_dccp_setsockopt
,
1140 .compat_getsockopt
= compat_dccp_getsockopt
,
1144 static struct inet6_protocol dccp_v6_protocol
= {
1145 .handler
= dccp_v6_rcv
,
1146 .err_handler
= dccp_v6_err
,
1147 .flags
= INET6_PROTO_NOPOLICY
| INET6_PROTO_FINAL
,
1150 static struct proto_ops inet6_dccp_ops
= {
1152 .owner
= THIS_MODULE
,
1153 .release
= inet6_release
,
1155 .connect
= inet_stream_connect
,
1156 .socketpair
= sock_no_socketpair
,
1157 .accept
= inet_accept
,
1158 .getname
= inet6_getname
,
1160 .ioctl
= inet6_ioctl
,
1161 .listen
= inet_dccp_listen
,
1162 .shutdown
= inet_shutdown
,
1163 .setsockopt
= sock_common_setsockopt
,
1164 .getsockopt
= sock_common_getsockopt
,
1165 .sendmsg
= inet_sendmsg
,
1166 .recvmsg
= sock_common_recvmsg
,
1167 .mmap
= sock_no_mmap
,
1168 .sendpage
= sock_no_sendpage
,
1169 #ifdef CONFIG_COMPAT
1170 .compat_setsockopt
= compat_sock_common_setsockopt
,
1171 .compat_getsockopt
= compat_sock_common_getsockopt
,
1175 static struct inet_protosw dccp_v6_protosw
= {
1177 .protocol
= IPPROTO_DCCP
,
1178 .prot
= &dccp_v6_prot
,
1179 .ops
= &inet6_dccp_ops
,
1181 .flags
= INET_PROTOSW_ICSK
,
1184 static int dccp_v6_init_net(struct net
*net
)
1188 err
= inet_ctl_sock_create(&net
->dccp
.v6_ctl_sk
, PF_INET6
,
1189 SOCK_DCCP
, IPPROTO_DCCP
, net
);
1193 static void dccp_v6_exit_net(struct net
*net
)
1195 inet_ctl_sock_destroy(net
->dccp
.v6_ctl_sk
);
1198 static struct pernet_operations dccp_v6_ops
= {
1199 .init
= dccp_v6_init_net
,
1200 .exit
= dccp_v6_exit_net
,
1203 static int __init
dccp_v6_init(void)
1205 int err
= proto_register(&dccp_v6_prot
, 1);
1210 err
= inet6_add_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1212 goto out_unregister_proto
;
1214 inet6_register_protosw(&dccp_v6_protosw
);
1216 err
= register_pernet_subsys(&dccp_v6_ops
);
1218 goto out_destroy_ctl_sock
;
1222 out_destroy_ctl_sock
:
1223 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1224 inet6_unregister_protosw(&dccp_v6_protosw
);
1225 out_unregister_proto
:
1226 proto_unregister(&dccp_v6_prot
);
1230 static void __exit
dccp_v6_exit(void)
1232 unregister_pernet_subsys(&dccp_v6_ops
);
1233 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1234 inet6_unregister_protosw(&dccp_v6_protosw
);
1235 proto_unregister(&dccp_v6_prot
);
1238 module_init(dccp_v6_init
);
1239 module_exit(dccp_v6_exit
);
1242 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1243 * values directly, Also cover the case where the protocol is not specified,
1244 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1246 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 33, 6);
1247 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 0, 6);
1248 MODULE_LICENSE("GPL");
1249 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1250 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");