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 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
);
105 sk
= inet6_lookup(net
, &dccp_hashinfo
,
106 &hdr
->daddr
, dh
->dccph_dport
,
107 &hdr
->saddr
, dh
->dccph_sport
, inet6_iif(skb
));
110 ICMP6_INC_STATS_BH(__in6_dev_get(skb
->dev
), ICMP6_MIB_INERRORS
);
114 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
115 inet_twsk_put(inet_twsk(sk
));
120 if (sock_owned_by_user(sk
))
121 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
123 if (sk
->sk_state
== DCCP_CLOSED
)
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
);
136 if (type
== ICMPV6_PKT_TOOBIG
) {
137 struct dst_entry
*dst
= NULL
;
139 if (sock_owned_by_user(sk
))
141 if ((1 << sk
->sk_state
) & (DCCPF_LISTEN
| DCCPF_CLOSED
))
144 /* icmp should have updated the destination cache entry */
145 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
147 struct inet_sock
*inet
= inet_sk(sk
);
150 /* BUGGG_FUTURE: Again, it is not clear how
151 to handle rthdr case. Ignore this complexity
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
);
165 sk
->sk_err_soft
= -err
;
169 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
171 sk
->sk_err_soft
= -err
;
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 */
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
;
190 if (sock_owned_by_user(sk
))
193 req
= inet6_csk_search_req(sk
, &prev
, dh
->dccph_dport
,
194 &hdr
->daddr
, &hdr
->saddr
,
200 * ICMPs are not backlogged, hence we cannot get an established
203 WARN_ON(req
->sk
!= NULL
);
205 if (seq
!= dccp_rsk(req
)->dreq_iss
) {
206 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
210 inet_csk_reqsk_queue_drop(sk
, req
, prev
);
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
);
220 * Wake people up to see the error
221 * (see connect in sock.c)
223 sk
->sk_error_report(sk
);
226 sk
->sk_err_soft
= err
;
230 if (!sock_owned_by_user(sk
) && np
->recverr
) {
232 sk
->sk_error_report(sk
);
234 sk
->sk_err_soft
= err
;
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
);
247 struct ipv6_txoptions
*opt
= NULL
;
248 struct in6_addr
*final_p
= NULL
, final
;
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;
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
);
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
);
273 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
278 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
280 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
284 skb
= dccp_make_response(sk
, dst
, req
);
286 struct dccp_hdr
*dh
= dccp_hdr(skb
);
288 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
,
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
);
297 if (opt
!= NULL
&& opt
!= np
->opt
)
298 sock_kfree_s(sk
, opt
, opt
->tot_len
);
303 static void dccp_v6_reqsk_destructor(struct request_sock
*req
)
305 dccp_feat_list_purge(&dccp_rsk(req
)->dreq_featneg
);
306 if (inet6_rsk(req
)->pktopts
!= NULL
)
307 kfree_skb(inet6_rsk(req
)->pktopts
);
310 static void dccp_v6_ctl_send_reset(struct sock
*sk
, struct sk_buff
*rxskb
)
312 struct ipv6hdr
*rxip6h
;
315 struct net
*net
= dev_net(rxskb
->dst
->dev
);
316 struct sock
*ctl_sk
= net
->dccp
.v6_ctl_sk
;
318 if (dccp_hdr(rxskb
)->dccph_type
== DCCP_PKT_RESET
)
321 if (!ipv6_unicast_destination(rxskb
))
324 skb
= dccp_ctl_make_reset(ctl_sk
, rxskb
);
328 rxip6h
= ipv6_hdr(rxskb
);
329 dccp_hdr(skb
)->dccph_checksum
= dccp_v6_csum_finish(skb
, &rxip6h
->saddr
,
332 memset(&fl
, 0, sizeof(fl
));
333 ipv6_addr_copy(&fl
.fl6_dst
, &rxip6h
->saddr
);
334 ipv6_addr_copy(&fl
.fl6_src
, &rxip6h
->daddr
);
336 fl
.proto
= IPPROTO_DCCP
;
337 fl
.oif
= inet6_iif(rxskb
);
338 fl
.fl_ip_dport
= dccp_hdr(skb
)->dccph_dport
;
339 fl
.fl_ip_sport
= dccp_hdr(skb
)->dccph_sport
;
340 security_skb_classify_flow(rxskb
, &fl
);
342 /* sk = NULL, but it is safe for now. RST socket required. */
343 if (!ip6_dst_lookup(ctl_sk
, &skb
->dst
, &fl
)) {
344 if (xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) >= 0) {
345 ip6_xmit(ctl_sk
, skb
, &fl
, NULL
, 0);
346 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS
);
347 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS
);
355 static struct request_sock_ops dccp6_request_sock_ops
= {
357 .obj_size
= sizeof(struct dccp6_request_sock
),
358 .rtx_syn_ack
= dccp_v6_send_response
,
359 .send_ack
= dccp_reqsk_send_ack
,
360 .destructor
= dccp_v6_reqsk_destructor
,
361 .send_reset
= dccp_v6_ctl_send_reset
,
364 static struct sock
*dccp_v6_hnd_req(struct sock
*sk
,struct sk_buff
*skb
)
366 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
367 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
369 struct request_sock
**prev
;
370 /* Find possible connection requests. */
371 struct request_sock
*req
= inet6_csk_search_req(sk
, &prev
,
377 return dccp_check_req(sk
, skb
, req
, prev
);
379 nsk
= __inet6_lookup_established(sock_net(sk
), &dccp_hashinfo
,
380 &iph
->saddr
, dh
->dccph_sport
,
381 &iph
->daddr
, ntohs(dh
->dccph_dport
),
384 if (nsk
->sk_state
!= DCCP_TIME_WAIT
) {
388 inet_twsk_put(inet_twsk(nsk
));
395 static int dccp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
397 struct request_sock
*req
;
398 struct dccp_request_sock
*dreq
;
399 struct inet6_request_sock
*ireq6
;
400 struct ipv6_pinfo
*np
= inet6_sk(sk
);
401 const __be32 service
= dccp_hdr_request(skb
)->dccph_req_service
;
402 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
404 if (skb
->protocol
== htons(ETH_P_IP
))
405 return dccp_v4_conn_request(sk
, skb
);
407 if (!ipv6_unicast_destination(skb
))
408 return 0; /* discard, don't send a reset here */
410 if (dccp_bad_service_code(sk
, service
)) {
411 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_BAD_SERVICE_CODE
;
415 * There are no SYN attacks on IPv6, yet...
417 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_TOO_BUSY
;
418 if (inet_csk_reqsk_queue_is_full(sk
))
421 if (sk_acceptq_is_full(sk
) && inet_csk_reqsk_queue_young(sk
) > 1)
424 req
= inet6_reqsk_alloc(&dccp6_request_sock_ops
);
428 if (dccp_reqsk_init(req
, dccp_sk(sk
), skb
))
431 dreq
= dccp_rsk(req
);
432 if (dccp_parse_options(sk
, dreq
, skb
))
435 if (security_inet_conn_request(sk
, skb
, req
))
438 ireq6
= inet6_rsk(req
);
439 ipv6_addr_copy(&ireq6
->rmt_addr
, &ipv6_hdr(skb
)->saddr
);
440 ipv6_addr_copy(&ireq6
->loc_addr
, &ipv6_hdr(skb
)->daddr
);
442 if (ipv6_opt_accepted(sk
, skb
) ||
443 np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
||
444 np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
) {
445 atomic_inc(&skb
->users
);
446 ireq6
->pktopts
= skb
;
448 ireq6
->iif
= sk
->sk_bound_dev_if
;
450 /* So that link locals have meaning */
451 if (!sk
->sk_bound_dev_if
&&
452 ipv6_addr_type(&ireq6
->rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
453 ireq6
->iif
= inet6_iif(skb
);
456 * Step 3: Process LISTEN state
458 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
460 * In fact we defer setting S.GSR, S.SWL, S.SWH to
461 * dccp_create_openreq_child.
463 dreq
->dreq_isr
= dcb
->dccpd_seq
;
464 dreq
->dreq_iss
= dccp_v6_init_sequence(skb
);
465 dreq
->dreq_service
= service
;
467 if (dccp_v6_send_response(sk
, req
))
470 inet6_csk_reqsk_queue_hash_add(sk
, req
, DCCP_TIMEOUT_INIT
);
476 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
480 static struct sock
*dccp_v6_request_recv_sock(struct sock
*sk
,
482 struct request_sock
*req
,
483 struct dst_entry
*dst
)
485 struct inet6_request_sock
*ireq6
= inet6_rsk(req
);
486 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
487 struct inet_sock
*newinet
;
488 struct dccp_sock
*newdp
;
489 struct dccp6_sock
*newdp6
;
491 struct ipv6_txoptions
*opt
;
493 if (skb
->protocol
== htons(ETH_P_IP
)) {
497 newsk
= dccp_v4_request_recv_sock(sk
, skb
, req
, dst
);
501 newdp6
= (struct dccp6_sock
*)newsk
;
502 newdp
= dccp_sk(newsk
);
503 newinet
= inet_sk(newsk
);
504 newinet
->pinet6
= &newdp6
->inet6
;
505 newnp
= inet6_sk(newsk
);
507 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
509 ipv6_addr_set(&newnp
->daddr
, 0, 0, htonl(0x0000FFFF),
512 ipv6_addr_set(&newnp
->saddr
, 0, 0, htonl(0x0000FFFF),
515 ipv6_addr_copy(&newnp
->rcv_saddr
, &newnp
->saddr
);
517 inet_csk(newsk
)->icsk_af_ops
= &dccp_ipv6_mapped
;
518 newsk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
519 newnp
->pktoptions
= NULL
;
521 newnp
->mcast_oif
= inet6_iif(skb
);
522 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
525 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
526 * here, dccp_create_openreq_child now does this for us, see the comment in
527 * that function for the gory details. -acme
530 /* It is tricky place. Until this moment IPv4 tcp
531 worked with IPv6 icsk.icsk_af_ops.
534 dccp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
541 if (sk_acceptq_is_full(sk
))
545 struct in6_addr
*final_p
= NULL
, final
;
548 memset(&fl
, 0, sizeof(fl
));
549 fl
.proto
= IPPROTO_DCCP
;
550 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
551 if (opt
!= NULL
&& opt
->srcrt
!= NULL
) {
552 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)opt
->srcrt
;
554 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
555 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
558 ipv6_addr_copy(&fl
.fl6_src
, &ireq6
->loc_addr
);
559 fl
.oif
= sk
->sk_bound_dev_if
;
560 fl
.fl_ip_dport
= inet_rsk(req
)->rmt_port
;
561 fl
.fl_ip_sport
= inet_sk(sk
)->sport
;
562 security_sk_classify_flow(sk
, &fl
);
564 if (ip6_dst_lookup(sk
, &dst
, &fl
))
568 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
570 if ((xfrm_lookup(&dst
, &fl
, sk
, 0)) < 0)
574 newsk
= dccp_create_openreq_child(sk
, req
, skb
);
579 * No need to charge this sock to the relevant IPv6 refcnt debug socks
580 * count here, dccp_create_openreq_child now does this for us, see the
581 * comment in that function for the gory details. -acme
584 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
585 newsk
->sk_route_caps
= dst
->dev
->features
& ~(NETIF_F_IP_CSUM
|
587 newdp6
= (struct dccp6_sock
*)newsk
;
588 newinet
= inet_sk(newsk
);
589 newinet
->pinet6
= &newdp6
->inet6
;
590 newdp
= dccp_sk(newsk
);
591 newnp
= inet6_sk(newsk
);
593 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
595 ipv6_addr_copy(&newnp
->daddr
, &ireq6
->rmt_addr
);
596 ipv6_addr_copy(&newnp
->saddr
, &ireq6
->loc_addr
);
597 ipv6_addr_copy(&newnp
->rcv_saddr
, &ireq6
->loc_addr
);
598 newsk
->sk_bound_dev_if
= ireq6
->iif
;
600 /* Now IPv6 options...
602 First: no IPv4 options.
607 newnp
->rxopt
.all
= np
->rxopt
.all
;
609 /* Clone pktoptions received with SYN */
610 newnp
->pktoptions
= NULL
;
611 if (ireq6
->pktopts
!= NULL
) {
612 newnp
->pktoptions
= skb_clone(ireq6
->pktopts
, GFP_ATOMIC
);
613 kfree_skb(ireq6
->pktopts
);
614 ireq6
->pktopts
= NULL
;
615 if (newnp
->pktoptions
)
616 skb_set_owner_r(newnp
->pktoptions
, newsk
);
619 newnp
->mcast_oif
= inet6_iif(skb
);
620 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
623 * Clone native IPv6 options from listening socket (if any)
625 * Yes, keeping reference count would be much more clever, but we make
626 * one more one thing there: reattach optmem to newsk.
629 newnp
->opt
= ipv6_dup_options(newsk
, opt
);
631 sock_kfree_s(sk
, opt
, opt
->tot_len
);
634 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
635 if (newnp
->opt
!= NULL
)
636 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
637 newnp
->opt
->opt_flen
);
639 dccp_sync_mss(newsk
, dst_mtu(dst
));
641 newinet
->daddr
= newinet
->saddr
= newinet
->rcv_saddr
= LOOPBACK4_IPV6
;
644 __inet_inherit_port(sk
, newsk
);
649 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
651 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
652 if (opt
!= NULL
&& opt
!= np
->opt
)
653 sock_kfree_s(sk
, opt
, opt
->tot_len
);
658 /* The socket must have it's spinlock held when we get
661 * We have a potential double-lock case here, so even when
662 * doing backlog processing we use the BH locking scheme.
663 * This is because we cannot sleep with the original spinlock
666 static int dccp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
668 struct ipv6_pinfo
*np
= inet6_sk(sk
);
669 struct sk_buff
*opt_skb
= NULL
;
671 /* Imagine: socket is IPv6. IPv4 packet arrives,
672 goes to IPv4 receive handler and backlogged.
673 From backlog it always goes here. Kerboom...
674 Fortunately, dccp_rcv_established and rcv_established
675 handle them correctly, but it is not case with
676 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
679 if (skb
->protocol
== htons(ETH_P_IP
))
680 return dccp_v4_do_rcv(sk
, skb
);
682 if (sk_filter(sk
, skb
))
686 * socket locking is here for SMP purposes as backlog rcv is currently
687 * called with bh processing disabled.
690 /* Do Stevens' IPV6_PKTOPTIONS.
692 Yes, guys, it is the only place in our code, where we
693 may make it not affecting IPv4.
694 The rest of code is protocol independent,
695 and I do not like idea to uglify IPv4.
697 Actually, all the idea behind IPV6_PKTOPTIONS
698 looks not very well thought. For now we latch
699 options, received in the last packet, enqueued
700 by tcp. Feel free to propose better solution.
705 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
706 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
708 opt_skb
= skb_clone(skb
, GFP_ATOMIC
);
710 if (sk
->sk_state
== DCCP_OPEN
) { /* Fast path */
711 if (dccp_rcv_established(sk
, skb
, dccp_hdr(skb
), skb
->len
))
714 /* XXX This is where we would goto ipv6_pktoptions. */
715 __kfree_skb(opt_skb
);
721 * Step 3: Process LISTEN state
722 * If S.state == LISTEN,
723 * If P.type == Request or P contains a valid Init Cookie option,
724 * (* Must scan the packet's options to check for Init
725 * Cookies. Only Init Cookies are processed here,
726 * however; other options are processed in Step 8. This
727 * scan need only be performed if the endpoint uses Init
729 * (* Generate a new socket and switch to that socket *)
730 * Set S := new socket for this port pair
732 * Choose S.ISS (initial seqno) or set from Init Cookies
733 * Initialize S.GAR := S.ISS
734 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
735 * Continue with S.state == RESPOND
736 * (* A Response packet will be generated in Step 11 *)
738 * Generate Reset(No Connection) unless P.type == Reset
739 * Drop packet and return
741 * NOTE: the check for the packet types is done in
742 * dccp_rcv_state_process
744 if (sk
->sk_state
== DCCP_LISTEN
) {
745 struct sock
*nsk
= dccp_v6_hnd_req(sk
, skb
);
750 * Queue it on the new socket if the new socket is active,
751 * otherwise we just shortcircuit this and continue with
755 if (dccp_child_process(sk
, nsk
, skb
))
758 __kfree_skb(opt_skb
);
763 if (dccp_rcv_state_process(sk
, skb
, dccp_hdr(skb
), skb
->len
))
766 /* XXX This is where we would goto ipv6_pktoptions. */
767 __kfree_skb(opt_skb
);
772 dccp_v6_ctl_send_reset(sk
, skb
);
775 __kfree_skb(opt_skb
);
780 static int dccp_v6_rcv(struct sk_buff
*skb
)
782 const struct dccp_hdr
*dh
;
786 /* Step 1: Check header basics */
788 if (dccp_invalid_packet(skb
))
791 /* Step 1: If header checksum is incorrect, drop packet and return. */
792 if (dccp_v6_csum_finish(skb
, &ipv6_hdr(skb
)->saddr
,
793 &ipv6_hdr(skb
)->daddr
)) {
794 DCCP_WARN("dropped packet with invalid checksum\n");
800 DCCP_SKB_CB(skb
)->dccpd_seq
= dccp_hdr_seq(dh
);
801 DCCP_SKB_CB(skb
)->dccpd_type
= dh
->dccph_type
;
803 if (dccp_packet_without_ack(skb
))
804 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= DCCP_PKT_WITHOUT_ACK_SEQ
;
806 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= dccp_hdr_ack_seq(skb
);
809 * Look up flow ID in table and get corresponding socket */
810 sk
= __inet6_lookup(dev_net(skb
->dst
->dev
), &dccp_hashinfo
,
811 &ipv6_hdr(skb
)->saddr
, dh
->dccph_sport
,
812 &ipv6_hdr(skb
)->daddr
, ntohs(dh
->dccph_dport
),
819 dccp_pr_debug("failed to look up flow ID in table and "
820 "get corresponding socket\n");
826 * ... or S.state == TIMEWAIT,
827 * Generate Reset(No Connection) unless P.type == Reset
828 * Drop packet and return
830 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
831 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
832 inet_twsk_put(inet_twsk(sk
));
837 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
838 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
839 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
841 min_cov
= dccp_sk(sk
)->dccps_pcrlen
;
842 if (dh
->dccph_cscov
&& (min_cov
== 0 || dh
->dccph_cscov
< min_cov
)) {
843 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
844 dh
->dccph_cscov
, min_cov
);
845 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
846 goto discard_and_relse
;
849 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
850 goto discard_and_relse
;
852 return sk_receive_skb(sk
, skb
, 1) ? -1 : 0;
855 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
860 * Generate Reset(No Connection) unless P.type == Reset
861 * Drop packet and return
863 if (dh
->dccph_type
!= DCCP_PKT_RESET
) {
864 DCCP_SKB_CB(skb
)->dccpd_reset_code
=
865 DCCP_RESET_CODE_NO_CONNECTION
;
866 dccp_v6_ctl_send_reset(sk
, skb
);
878 static int dccp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
881 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*)uaddr
;
882 struct inet_connection_sock
*icsk
= inet_csk(sk
);
883 struct inet_sock
*inet
= inet_sk(sk
);
884 struct ipv6_pinfo
*np
= inet6_sk(sk
);
885 struct dccp_sock
*dp
= dccp_sk(sk
);
886 struct in6_addr
*saddr
= NULL
, *final_p
= NULL
, final
;
888 struct dst_entry
*dst
;
892 dp
->dccps_role
= DCCP_ROLE_CLIENT
;
894 if (addr_len
< SIN6_LEN_RFC2133
)
897 if (usin
->sin6_family
!= AF_INET6
)
898 return -EAFNOSUPPORT
;
900 memset(&fl
, 0, sizeof(fl
));
903 fl
.fl6_flowlabel
= usin
->sin6_flowinfo
& IPV6_FLOWINFO_MASK
;
904 IP6_ECN_flow_init(fl
.fl6_flowlabel
);
905 if (fl
.fl6_flowlabel
& IPV6_FLOWLABEL_MASK
) {
906 struct ip6_flowlabel
*flowlabel
;
907 flowlabel
= fl6_sock_lookup(sk
, fl
.fl6_flowlabel
);
908 if (flowlabel
== NULL
)
910 ipv6_addr_copy(&usin
->sin6_addr
, &flowlabel
->dst
);
911 fl6_sock_release(flowlabel
);
915 * connect() to INADDR_ANY means loopback (BSD'ism).
917 if (ipv6_addr_any(&usin
->sin6_addr
))
918 usin
->sin6_addr
.s6_addr
[15] = 1;
920 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
922 if (addr_type
& IPV6_ADDR_MULTICAST
)
925 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
926 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
927 usin
->sin6_scope_id
) {
928 /* If interface is set while binding, indices
931 if (sk
->sk_bound_dev_if
&&
932 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
935 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
938 /* Connect to link-local address requires an interface */
939 if (!sk
->sk_bound_dev_if
)
943 ipv6_addr_copy(&np
->daddr
, &usin
->sin6_addr
);
944 np
->flow_label
= fl
.fl6_flowlabel
;
949 if (addr_type
== IPV6_ADDR_MAPPED
) {
950 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
951 struct sockaddr_in sin
;
953 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
955 if (__ipv6_only_sock(sk
))
958 sin
.sin_family
= AF_INET
;
959 sin
.sin_port
= usin
->sin6_port
;
960 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
962 icsk
->icsk_af_ops
= &dccp_ipv6_mapped
;
963 sk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
965 err
= dccp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
967 icsk
->icsk_ext_hdr_len
= exthdrlen
;
968 icsk
->icsk_af_ops
= &dccp_ipv6_af_ops
;
969 sk
->sk_backlog_rcv
= dccp_v6_do_rcv
;
972 ipv6_addr_set(&np
->saddr
, 0, 0, htonl(0x0000FFFF),
974 ipv6_addr_set(&np
->rcv_saddr
, 0, 0, htonl(0x0000FFFF),
981 if (!ipv6_addr_any(&np
->rcv_saddr
))
982 saddr
= &np
->rcv_saddr
;
984 fl
.proto
= IPPROTO_DCCP
;
985 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
986 ipv6_addr_copy(&fl
.fl6_src
, saddr
? saddr
: &np
->saddr
);
987 fl
.oif
= sk
->sk_bound_dev_if
;
988 fl
.fl_ip_dport
= usin
->sin6_port
;
989 fl
.fl_ip_sport
= inet
->sport
;
990 security_sk_classify_flow(sk
, &fl
);
992 if (np
->opt
!= NULL
&& np
->opt
->srcrt
!= NULL
) {
993 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)np
->opt
->srcrt
;
995 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
996 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
1000 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
1005 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
1007 err
= __xfrm_lookup(&dst
, &fl
, sk
, XFRM_LOOKUP_WAIT
);
1009 if (err
== -EREMOTE
)
1010 err
= ip6_dst_blackhole(sk
, &dst
, &fl
);
1015 if (saddr
== NULL
) {
1016 saddr
= &fl
.fl6_src
;
1017 ipv6_addr_copy(&np
->rcv_saddr
, saddr
);
1020 /* set the source address */
1021 ipv6_addr_copy(&np
->saddr
, saddr
);
1022 inet
->rcv_saddr
= LOOPBACK4_IPV6
;
1024 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
1026 icsk
->icsk_ext_hdr_len
= 0;
1027 if (np
->opt
!= NULL
)
1028 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
1029 np
->opt
->opt_nflen
);
1031 inet
->dport
= usin
->sin6_port
;
1033 dccp_set_state(sk
, DCCP_REQUESTING
);
1034 err
= inet6_hash_connect(&dccp_death_row
, sk
);
1038 dp
->dccps_iss
= secure_dccpv6_sequence_number(np
->saddr
.s6_addr32
,
1039 np
->daddr
.s6_addr32
,
1040 inet
->sport
, inet
->dport
);
1041 err
= dccp_connect(sk
);
1048 dccp_set_state(sk
, DCCP_CLOSED
);
1052 sk
->sk_route_caps
= 0;
1056 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops
= {
1057 .queue_xmit
= inet6_csk_xmit
,
1058 .send_check
= dccp_v6_send_check
,
1059 .rebuild_header
= inet6_sk_rebuild_header
,
1060 .conn_request
= dccp_v6_conn_request
,
1061 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1062 .net_header_len
= sizeof(struct ipv6hdr
),
1063 .setsockopt
= ipv6_setsockopt
,
1064 .getsockopt
= ipv6_getsockopt
,
1065 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1066 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1067 .bind_conflict
= inet6_csk_bind_conflict
,
1068 #ifdef CONFIG_COMPAT
1069 .compat_setsockopt
= compat_ipv6_setsockopt
,
1070 .compat_getsockopt
= compat_ipv6_getsockopt
,
1075 * DCCP over IPv4 via INET6 API
1077 static struct inet_connection_sock_af_ops dccp_ipv6_mapped
= {
1078 .queue_xmit
= ip_queue_xmit
,
1079 .send_check
= dccp_v4_send_check
,
1080 .rebuild_header
= inet_sk_rebuild_header
,
1081 .conn_request
= dccp_v6_conn_request
,
1082 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1083 .net_header_len
= sizeof(struct iphdr
),
1084 .setsockopt
= ipv6_setsockopt
,
1085 .getsockopt
= ipv6_getsockopt
,
1086 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1087 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1088 #ifdef CONFIG_COMPAT
1089 .compat_setsockopt
= compat_ipv6_setsockopt
,
1090 .compat_getsockopt
= compat_ipv6_getsockopt
,
1094 /* NOTE: A lot of things set to zero explicitly by call to
1095 * sk_alloc() so need not be done here.
1097 static int dccp_v6_init_sock(struct sock
*sk
)
1099 static __u8 dccp_v6_ctl_sock_initialized
;
1100 int err
= dccp_init_sock(sk
, dccp_v6_ctl_sock_initialized
);
1103 if (unlikely(!dccp_v6_ctl_sock_initialized
))
1104 dccp_v6_ctl_sock_initialized
= 1;
1105 inet_csk(sk
)->icsk_af_ops
= &dccp_ipv6_af_ops
;
1111 static void dccp_v6_destroy_sock(struct sock
*sk
)
1113 dccp_destroy_sock(sk
);
1114 inet6_destroy_sock(sk
);
1117 static struct timewait_sock_ops dccp6_timewait_sock_ops
= {
1118 .twsk_obj_size
= sizeof(struct dccp6_timewait_sock
),
1121 static struct proto dccp_v6_prot
= {
1123 .owner
= THIS_MODULE
,
1124 .close
= dccp_close
,
1125 .connect
= dccp_v6_connect
,
1126 .disconnect
= dccp_disconnect
,
1127 .ioctl
= dccp_ioctl
,
1128 .init
= dccp_v6_init_sock
,
1129 .setsockopt
= dccp_setsockopt
,
1130 .getsockopt
= dccp_getsockopt
,
1131 .sendmsg
= dccp_sendmsg
,
1132 .recvmsg
= dccp_recvmsg
,
1133 .backlog_rcv
= dccp_v6_do_rcv
,
1134 .hash
= dccp_v6_hash
,
1135 .unhash
= inet_unhash
,
1136 .accept
= inet_csk_accept
,
1137 .get_port
= inet_csk_get_port
,
1138 .shutdown
= dccp_shutdown
,
1139 .destroy
= dccp_v6_destroy_sock
,
1140 .orphan_count
= &dccp_orphan_count
,
1141 .max_header
= MAX_DCCP_HEADER
,
1142 .obj_size
= sizeof(struct dccp6_sock
),
1143 .rsk_prot
= &dccp6_request_sock_ops
,
1144 .twsk_prot
= &dccp6_timewait_sock_ops
,
1145 .h
.hashinfo
= &dccp_hashinfo
,
1146 #ifdef CONFIG_COMPAT
1147 .compat_setsockopt
= compat_dccp_setsockopt
,
1148 .compat_getsockopt
= compat_dccp_getsockopt
,
1152 static struct inet6_protocol dccp_v6_protocol
= {
1153 .handler
= dccp_v6_rcv
,
1154 .err_handler
= dccp_v6_err
,
1155 .flags
= INET6_PROTO_NOPOLICY
| INET6_PROTO_FINAL
,
1158 static struct proto_ops inet6_dccp_ops
= {
1160 .owner
= THIS_MODULE
,
1161 .release
= inet6_release
,
1163 .connect
= inet_stream_connect
,
1164 .socketpair
= sock_no_socketpair
,
1165 .accept
= inet_accept
,
1166 .getname
= inet6_getname
,
1168 .ioctl
= inet6_ioctl
,
1169 .listen
= inet_dccp_listen
,
1170 .shutdown
= inet_shutdown
,
1171 .setsockopt
= sock_common_setsockopt
,
1172 .getsockopt
= sock_common_getsockopt
,
1173 .sendmsg
= inet_sendmsg
,
1174 .recvmsg
= sock_common_recvmsg
,
1175 .mmap
= sock_no_mmap
,
1176 .sendpage
= sock_no_sendpage
,
1177 #ifdef CONFIG_COMPAT
1178 .compat_setsockopt
= compat_sock_common_setsockopt
,
1179 .compat_getsockopt
= compat_sock_common_getsockopt
,
1183 static struct inet_protosw dccp_v6_protosw
= {
1185 .protocol
= IPPROTO_DCCP
,
1186 .prot
= &dccp_v6_prot
,
1187 .ops
= &inet6_dccp_ops
,
1189 .flags
= INET_PROTOSW_ICSK
,
1192 static int dccp_v6_init_net(struct net
*net
)
1196 err
= inet_ctl_sock_create(&net
->dccp
.v6_ctl_sk
, PF_INET6
,
1197 SOCK_DCCP
, IPPROTO_DCCP
, net
);
1201 static void dccp_v6_exit_net(struct net
*net
)
1203 inet_ctl_sock_destroy(net
->dccp
.v6_ctl_sk
);
1206 static struct pernet_operations dccp_v6_ops
= {
1207 .init
= dccp_v6_init_net
,
1208 .exit
= dccp_v6_exit_net
,
1211 static int __init
dccp_v6_init(void)
1213 int err
= proto_register(&dccp_v6_prot
, 1);
1218 err
= inet6_add_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1220 goto out_unregister_proto
;
1222 inet6_register_protosw(&dccp_v6_protosw
);
1224 err
= register_pernet_subsys(&dccp_v6_ops
);
1226 goto out_destroy_ctl_sock
;
1230 out_destroy_ctl_sock
:
1231 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1232 inet6_unregister_protosw(&dccp_v6_protosw
);
1233 out_unregister_proto
:
1234 proto_unregister(&dccp_v6_prot
);
1238 static void __exit
dccp_v6_exit(void)
1240 unregister_pernet_subsys(&dccp_v6_ops
);
1241 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1242 inet6_unregister_protosw(&dccp_v6_protosw
);
1243 proto_unregister(&dccp_v6_prot
);
1246 module_init(dccp_v6_init
);
1247 module_exit(dccp_v6_exit
);
1250 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1251 * values directly, Also cover the case where the protocol is not specified,
1252 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1254 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 33, 6);
1255 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 0, 6);
1256 MODULE_LICENSE("GPL");
1257 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1258 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");