3 * Linux INET6 implementation
5 * Based on net/dccp6/ipv6.c
7 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
15 #include <linux/module.h>
16 #include <linux/random.h>
17 #include <linux/slab.h>
18 #include <linux/xfrm.h>
20 #include <net/addrconf.h>
21 #include <net/inet_common.h>
22 #include <net/inet_hashtables.h>
23 #include <net/inet_sock.h>
24 #include <net/inet6_connection_sock.h>
25 #include <net/inet6_hashtables.h>
26 #include <net/ip6_route.h>
28 #include <net/protocol.h>
29 #include <net/transp_v6.h>
30 #include <net/ip6_checksum.h>
32 #include <net/secure_seq.h>
38 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
40 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped
;
41 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops
;
43 static void dccp_v6_hash(struct sock
*sk
)
45 if (sk
->sk_state
!= DCCP_CLOSED
) {
46 if (inet_csk(sk
)->icsk_af_ops
== &dccp_ipv6_mapped
) {
51 __inet6_hash(sk
, NULL
);
56 /* add pseudo-header to DCCP checksum stored in skb->csum */
57 static inline __sum16
dccp_v6_csum_finish(struct sk_buff
*skb
,
58 const struct in6_addr
*saddr
,
59 const struct in6_addr
*daddr
)
61 return csum_ipv6_magic(saddr
, daddr
, skb
->len
, IPPROTO_DCCP
, skb
->csum
);
64 static inline void dccp_v6_send_check(struct sock
*sk
, struct sk_buff
*skb
)
66 struct ipv6_pinfo
*np
= inet6_sk(sk
);
67 struct dccp_hdr
*dh
= dccp_hdr(skb
);
69 dccp_csum_outgoing(skb
);
70 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
, &np
->saddr
, &sk
->sk_v6_daddr
);
73 static inline __u64
dccp_v6_init_sequence(struct sk_buff
*skb
)
75 return secure_dccpv6_sequence_number(ipv6_hdr(skb
)->daddr
.s6_addr32
,
76 ipv6_hdr(skb
)->saddr
.s6_addr32
,
77 dccp_hdr(skb
)->dccph_dport
,
78 dccp_hdr(skb
)->dccph_sport
);
82 static void dccp_v6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
83 u8 type
, u8 code
, int offset
, __be32 info
)
85 const struct ipv6hdr
*hdr
= (const struct ipv6hdr
*)skb
->data
;
86 const struct dccp_hdr
*dh
= (struct dccp_hdr
*)(skb
->data
+ offset
);
88 struct ipv6_pinfo
*np
;
92 struct net
*net
= dev_net(skb
->dev
);
94 if (skb
->len
< offset
+ sizeof(*dh
) ||
95 skb
->len
< offset
+ __dccp_basic_hdr_len(dh
)) {
96 ICMP6_INC_STATS_BH(net
, __in6_dev_get(skb
->dev
),
101 sk
= inet6_lookup(net
, &dccp_hashinfo
,
102 &hdr
->daddr
, dh
->dccph_dport
,
103 &hdr
->saddr
, dh
->dccph_sport
, inet6_iif(skb
));
106 ICMP6_INC_STATS_BH(net
, __in6_dev_get(skb
->dev
),
111 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
112 inet_twsk_put(inet_twsk(sk
));
117 if (sock_owned_by_user(sk
))
118 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
120 if (sk
->sk_state
== DCCP_CLOSED
)
124 seq
= dccp_hdr_seq(dh
);
125 if ((1 << sk
->sk_state
) & ~(DCCPF_REQUESTING
| DCCPF_LISTEN
) &&
126 !between48(seq
, dp
->dccps_awl
, dp
->dccps_awh
)) {
127 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
133 if (type
== NDISC_REDIRECT
) {
134 struct dst_entry
*dst
= __sk_dst_check(sk
, np
->dst_cookie
);
137 dst
->ops
->redirect(dst
, sk
, skb
);
141 if (type
== ICMPV6_PKT_TOOBIG
) {
142 struct dst_entry
*dst
= NULL
;
144 if (!ip6_sk_accept_pmtu(sk
))
147 if (sock_owned_by_user(sk
))
149 if ((1 << sk
->sk_state
) & (DCCPF_LISTEN
| DCCPF_CLOSED
))
152 dst
= inet6_csk_update_pmtu(sk
, ntohl(info
));
156 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
))
157 dccp_sync_mss(sk
, dst_mtu(dst
));
161 icmpv6_err_convert(type
, code
, &err
);
163 /* Might be for an request_sock */
164 switch (sk
->sk_state
) {
165 struct request_sock
*req
, **prev
;
167 if (sock_owned_by_user(sk
))
170 req
= inet6_csk_search_req(sk
, &prev
, dh
->dccph_dport
,
171 &hdr
->daddr
, &hdr
->saddr
,
177 * ICMPs are not backlogged, hence we cannot get an established
180 WARN_ON(req
->sk
!= NULL
);
182 if (!between48(seq
, dccp_rsk(req
)->dreq_iss
,
183 dccp_rsk(req
)->dreq_gss
)) {
184 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
188 inet_csk_reqsk_queue_drop(sk
, req
, prev
);
191 case DCCP_REQUESTING
:
192 case DCCP_RESPOND
: /* Cannot happen.
193 It can, it SYNs are crossed. --ANK */
194 if (!sock_owned_by_user(sk
)) {
195 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
198 * Wake people up to see the error
199 * (see connect in sock.c)
201 sk
->sk_error_report(sk
);
204 sk
->sk_err_soft
= err
;
208 if (!sock_owned_by_user(sk
) && np
->recverr
) {
210 sk
->sk_error_report(sk
);
212 sk
->sk_err_soft
= err
;
220 static int dccp_v6_send_response(struct sock
*sk
, struct request_sock
*req
)
222 struct inet_request_sock
*ireq
= inet_rsk(req
);
223 struct ipv6_pinfo
*np
= inet6_sk(sk
);
225 struct in6_addr
*final_p
, final
;
228 struct dst_entry
*dst
;
230 memset(&fl6
, 0, sizeof(fl6
));
231 fl6
.flowi6_proto
= IPPROTO_DCCP
;
232 fl6
.daddr
= ireq
->ir_v6_rmt_addr
;
233 fl6
.saddr
= ireq
->ir_v6_loc_addr
;
235 fl6
.flowi6_oif
= ireq
->ir_iif
;
236 fl6
.fl6_dport
= ireq
->ir_rmt_port
;
237 fl6
.fl6_sport
= htons(ireq
->ir_num
);
238 security_req_classify_flow(req
, flowi6_to_flowi(&fl6
));
241 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
243 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
250 skb
= dccp_make_response(sk
, dst
, req
);
252 struct dccp_hdr
*dh
= dccp_hdr(skb
);
254 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
,
255 &ireq
->ir_v6_loc_addr
,
256 &ireq
->ir_v6_rmt_addr
);
257 fl6
.daddr
= ireq
->ir_v6_rmt_addr
;
258 err
= ip6_xmit(sk
, skb
, &fl6
, np
->opt
, np
->tclass
);
259 err
= net_xmit_eval(err
);
267 static void dccp_v6_reqsk_destructor(struct request_sock
*req
)
269 dccp_feat_list_purge(&dccp_rsk(req
)->dreq_featneg
);
270 kfree_skb(inet_rsk(req
)->pktopts
);
273 static void dccp_v6_ctl_send_reset(struct sock
*sk
, struct sk_buff
*rxskb
)
275 const struct ipv6hdr
*rxip6h
;
278 struct net
*net
= dev_net(skb_dst(rxskb
)->dev
);
279 struct sock
*ctl_sk
= net
->dccp
.v6_ctl_sk
;
280 struct dst_entry
*dst
;
282 if (dccp_hdr(rxskb
)->dccph_type
== DCCP_PKT_RESET
)
285 if (!ipv6_unicast_destination(rxskb
))
288 skb
= dccp_ctl_make_reset(ctl_sk
, rxskb
);
292 rxip6h
= ipv6_hdr(rxskb
);
293 dccp_hdr(skb
)->dccph_checksum
= dccp_v6_csum_finish(skb
, &rxip6h
->saddr
,
296 memset(&fl6
, 0, sizeof(fl6
));
297 fl6
.daddr
= rxip6h
->saddr
;
298 fl6
.saddr
= rxip6h
->daddr
;
300 fl6
.flowi6_proto
= IPPROTO_DCCP
;
301 fl6
.flowi6_oif
= inet6_iif(rxskb
);
302 fl6
.fl6_dport
= dccp_hdr(skb
)->dccph_dport
;
303 fl6
.fl6_sport
= dccp_hdr(skb
)->dccph_sport
;
304 security_skb_classify_flow(rxskb
, flowi6_to_flowi(&fl6
));
306 /* sk = NULL, but it is safe for now. RST socket required. */
307 dst
= ip6_dst_lookup_flow(ctl_sk
, &fl6
, NULL
);
309 skb_dst_set(skb
, dst
);
310 ip6_xmit(ctl_sk
, skb
, &fl6
, NULL
, 0);
311 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS
);
312 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS
);
319 static struct request_sock_ops dccp6_request_sock_ops
= {
321 .obj_size
= sizeof(struct dccp6_request_sock
),
322 .rtx_syn_ack
= dccp_v6_send_response
,
323 .send_ack
= dccp_reqsk_send_ack
,
324 .destructor
= dccp_v6_reqsk_destructor
,
325 .send_reset
= dccp_v6_ctl_send_reset
,
326 .syn_ack_timeout
= dccp_syn_ack_timeout
,
329 static struct sock
*dccp_v6_hnd_req(struct sock
*sk
,struct sk_buff
*skb
)
331 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
332 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
334 struct request_sock
**prev
;
335 /* Find possible connection requests. */
336 struct request_sock
*req
= inet6_csk_search_req(sk
, &prev
,
342 return dccp_check_req(sk
, skb
, req
, prev
);
344 nsk
= __inet6_lookup_established(sock_net(sk
), &dccp_hashinfo
,
345 &iph
->saddr
, dh
->dccph_sport
,
346 &iph
->daddr
, ntohs(dh
->dccph_dport
),
349 if (nsk
->sk_state
!= DCCP_TIME_WAIT
) {
353 inet_twsk_put(inet_twsk(nsk
));
360 static int dccp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
362 struct request_sock
*req
;
363 struct dccp_request_sock
*dreq
;
364 struct inet_request_sock
*ireq
;
365 struct ipv6_pinfo
*np
= inet6_sk(sk
);
366 const __be32 service
= dccp_hdr_request(skb
)->dccph_req_service
;
367 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
369 if (skb
->protocol
== htons(ETH_P_IP
))
370 return dccp_v4_conn_request(sk
, skb
);
372 if (!ipv6_unicast_destination(skb
))
373 return 0; /* discard, don't send a reset here */
375 if (dccp_bad_service_code(sk
, service
)) {
376 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_BAD_SERVICE_CODE
;
380 * There are no SYN attacks on IPv6, yet...
382 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_TOO_BUSY
;
383 if (inet_csk_reqsk_queue_is_full(sk
))
386 if (sk_acceptq_is_full(sk
) && inet_csk_reqsk_queue_young(sk
) > 1)
389 req
= inet_reqsk_alloc(&dccp6_request_sock_ops
);
393 if (dccp_reqsk_init(req
, dccp_sk(sk
), skb
))
396 dreq
= dccp_rsk(req
);
397 if (dccp_parse_options(sk
, dreq
, skb
))
400 if (security_inet_conn_request(sk
, skb
, req
))
403 ireq
= inet_rsk(req
);
404 ireq
->ir_v6_rmt_addr
= ipv6_hdr(skb
)->saddr
;
405 ireq
->ir_v6_loc_addr
= ipv6_hdr(skb
)->daddr
;
406 write_pnet(&ireq
->ireq_net
, sock_net(sk
));
407 ireq
->ireq_family
= AF_INET6
;
409 if (ipv6_opt_accepted(sk
, skb
, IP6CB(skb
)) ||
410 np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
||
411 np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
) {
412 atomic_inc(&skb
->users
);
415 ireq
->ir_iif
= sk
->sk_bound_dev_if
;
417 /* So that link locals have meaning */
418 if (!sk
->sk_bound_dev_if
&&
419 ipv6_addr_type(&ireq
->ir_v6_rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
420 ireq
->ir_iif
= inet6_iif(skb
);
423 * Step 3: Process LISTEN state
425 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
427 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
429 dreq
->dreq_isr
= dcb
->dccpd_seq
;
430 dreq
->dreq_gsr
= dreq
->dreq_isr
;
431 dreq
->dreq_iss
= dccp_v6_init_sequence(skb
);
432 dreq
->dreq_gss
= dreq
->dreq_iss
;
433 dreq
->dreq_service
= service
;
435 if (dccp_v6_send_response(sk
, req
))
438 inet6_csk_reqsk_queue_hash_add(sk
, req
, DCCP_TIMEOUT_INIT
);
444 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
448 static struct sock
*dccp_v6_request_recv_sock(struct sock
*sk
,
450 struct request_sock
*req
,
451 struct dst_entry
*dst
)
453 struct inet_request_sock
*ireq
= inet_rsk(req
);
454 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
455 struct inet_sock
*newinet
;
456 struct dccp6_sock
*newdp6
;
459 if (skb
->protocol
== htons(ETH_P_IP
)) {
463 newsk
= dccp_v4_request_recv_sock(sk
, skb
, req
, dst
);
467 newdp6
= (struct dccp6_sock
*)newsk
;
468 newinet
= inet_sk(newsk
);
469 newinet
->pinet6
= &newdp6
->inet6
;
470 newnp
= inet6_sk(newsk
);
472 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
474 ipv6_addr_set_v4mapped(newinet
->inet_daddr
, &newsk
->sk_v6_daddr
);
476 ipv6_addr_set_v4mapped(newinet
->inet_saddr
, &newnp
->saddr
);
478 newsk
->sk_v6_rcv_saddr
= newnp
->saddr
;
480 inet_csk(newsk
)->icsk_af_ops
= &dccp_ipv6_mapped
;
481 newsk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
482 newnp
->pktoptions
= NULL
;
484 newnp
->mcast_oif
= inet6_iif(skb
);
485 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
488 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
489 * here, dccp_create_openreq_child now does this for us, see the comment in
490 * that function for the gory details. -acme
493 /* It is tricky place. Until this moment IPv4 tcp
494 worked with IPv6 icsk.icsk_af_ops.
497 dccp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
503 if (sk_acceptq_is_full(sk
))
507 struct in6_addr
*final_p
, final
;
510 memset(&fl6
, 0, sizeof(fl6
));
511 fl6
.flowi6_proto
= IPPROTO_DCCP
;
512 fl6
.daddr
= ireq
->ir_v6_rmt_addr
;
513 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
514 fl6
.saddr
= ireq
->ir_v6_loc_addr
;
515 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
516 fl6
.fl6_dport
= ireq
->ir_rmt_port
;
517 fl6
.fl6_sport
= htons(ireq
->ir_num
);
518 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
520 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
525 newsk
= dccp_create_openreq_child(sk
, req
, skb
);
530 * No need to charge this sock to the relevant IPv6 refcnt debug socks
531 * count here, dccp_create_openreq_child now does this for us, see the
532 * comment in that function for the gory details. -acme
535 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
536 newsk
->sk_route_caps
= dst
->dev
->features
& ~(NETIF_F_IP_CSUM
|
538 newdp6
= (struct dccp6_sock
*)newsk
;
539 newinet
= inet_sk(newsk
);
540 newinet
->pinet6
= &newdp6
->inet6
;
541 newnp
= inet6_sk(newsk
);
543 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
545 newsk
->sk_v6_daddr
= ireq
->ir_v6_rmt_addr
;
546 newnp
->saddr
= ireq
->ir_v6_loc_addr
;
547 newsk
->sk_v6_rcv_saddr
= ireq
->ir_v6_loc_addr
;
548 newsk
->sk_bound_dev_if
= ireq
->ir_iif
;
550 /* Now IPv6 options...
552 First: no IPv4 options.
554 newinet
->inet_opt
= NULL
;
557 newnp
->rxopt
.all
= np
->rxopt
.all
;
559 /* Clone pktoptions received with SYN */
560 newnp
->pktoptions
= NULL
;
561 if (ireq
->pktopts
!= NULL
) {
562 newnp
->pktoptions
= skb_clone(ireq
->pktopts
, GFP_ATOMIC
);
563 consume_skb(ireq
->pktopts
);
564 ireq
->pktopts
= NULL
;
565 if (newnp
->pktoptions
)
566 skb_set_owner_r(newnp
->pktoptions
, newsk
);
569 newnp
->mcast_oif
= inet6_iif(skb
);
570 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
573 * Clone native IPv6 options from listening socket (if any)
575 * Yes, keeping reference count would be much more clever, but we make
576 * one more one thing there: reattach optmem to newsk.
579 newnp
->opt
= ipv6_dup_options(newsk
, np
->opt
);
581 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
582 if (newnp
->opt
!= NULL
)
583 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
584 newnp
->opt
->opt_flen
);
586 dccp_sync_mss(newsk
, dst_mtu(dst
));
588 newinet
->inet_daddr
= newinet
->inet_saddr
= LOOPBACK4_IPV6
;
589 newinet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
591 if (__inet_inherit_port(sk
, newsk
) < 0) {
592 inet_csk_prepare_forced_close(newsk
);
596 __inet6_hash(newsk
, NULL
);
601 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
605 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
609 /* The socket must have it's spinlock held when we get
612 * We have a potential double-lock case here, so even when
613 * doing backlog processing we use the BH locking scheme.
614 * This is because we cannot sleep with the original spinlock
617 static int dccp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
619 struct ipv6_pinfo
*np
= inet6_sk(sk
);
620 struct sk_buff
*opt_skb
= NULL
;
622 /* Imagine: socket is IPv6. IPv4 packet arrives,
623 goes to IPv4 receive handler and backlogged.
624 From backlog it always goes here. Kerboom...
625 Fortunately, dccp_rcv_established and rcv_established
626 handle them correctly, but it is not case with
627 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
630 if (skb
->protocol
== htons(ETH_P_IP
))
631 return dccp_v4_do_rcv(sk
, skb
);
633 if (sk_filter(sk
, skb
))
637 * socket locking is here for SMP purposes as backlog rcv is currently
638 * called with bh processing disabled.
641 /* Do Stevens' IPV6_PKTOPTIONS.
643 Yes, guys, it is the only place in our code, where we
644 may make it not affecting IPv4.
645 The rest of code is protocol independent,
646 and I do not like idea to uglify IPv4.
648 Actually, all the idea behind IPV6_PKTOPTIONS
649 looks not very well thought. For now we latch
650 options, received in the last packet, enqueued
651 by tcp. Feel free to propose better solution.
656 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
657 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
659 opt_skb
= skb_clone(skb
, GFP_ATOMIC
);
661 if (sk
->sk_state
== DCCP_OPEN
) { /* Fast path */
662 if (dccp_rcv_established(sk
, skb
, dccp_hdr(skb
), skb
->len
))
665 /* XXX This is where we would goto ipv6_pktoptions. */
666 __kfree_skb(opt_skb
);
672 * Step 3: Process LISTEN state
673 * If S.state == LISTEN,
674 * If P.type == Request or P contains a valid Init Cookie option,
675 * (* Must scan the packet's options to check for Init
676 * Cookies. Only Init Cookies are processed here,
677 * however; other options are processed in Step 8. This
678 * scan need only be performed if the endpoint uses Init
680 * (* Generate a new socket and switch to that socket *)
681 * Set S := new socket for this port pair
683 * Choose S.ISS (initial seqno) or set from Init Cookies
684 * Initialize S.GAR := S.ISS
685 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
686 * Continue with S.state == RESPOND
687 * (* A Response packet will be generated in Step 11 *)
689 * Generate Reset(No Connection) unless P.type == Reset
690 * Drop packet and return
692 * NOTE: the check for the packet types is done in
693 * dccp_rcv_state_process
695 if (sk
->sk_state
== DCCP_LISTEN
) {
696 struct sock
*nsk
= dccp_v6_hnd_req(sk
, skb
);
701 * Queue it on the new socket if the new socket is active,
702 * otherwise we just shortcircuit this and continue with
706 if (dccp_child_process(sk
, nsk
, skb
))
709 __kfree_skb(opt_skb
);
714 if (dccp_rcv_state_process(sk
, skb
, dccp_hdr(skb
), skb
->len
))
717 /* XXX This is where we would goto ipv6_pktoptions. */
718 __kfree_skb(opt_skb
);
723 dccp_v6_ctl_send_reset(sk
, skb
);
726 __kfree_skb(opt_skb
);
731 static int dccp_v6_rcv(struct sk_buff
*skb
)
733 const struct dccp_hdr
*dh
;
737 /* Step 1: Check header basics */
739 if (dccp_invalid_packet(skb
))
742 /* Step 1: If header checksum is incorrect, drop packet and return. */
743 if (dccp_v6_csum_finish(skb
, &ipv6_hdr(skb
)->saddr
,
744 &ipv6_hdr(skb
)->daddr
)) {
745 DCCP_WARN("dropped packet with invalid checksum\n");
751 DCCP_SKB_CB(skb
)->dccpd_seq
= dccp_hdr_seq(dh
);
752 DCCP_SKB_CB(skb
)->dccpd_type
= dh
->dccph_type
;
754 if (dccp_packet_without_ack(skb
))
755 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= DCCP_PKT_WITHOUT_ACK_SEQ
;
757 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= dccp_hdr_ack_seq(skb
);
760 * Look up flow ID in table and get corresponding socket */
761 sk
= __inet6_lookup_skb(&dccp_hashinfo
, skb
,
762 dh
->dccph_sport
, dh
->dccph_dport
,
769 dccp_pr_debug("failed to look up flow ID in table and "
770 "get corresponding socket\n");
776 * ... or S.state == TIMEWAIT,
777 * Generate Reset(No Connection) unless P.type == Reset
778 * Drop packet and return
780 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
781 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
782 inet_twsk_put(inet_twsk(sk
));
787 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
788 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
789 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
791 min_cov
= dccp_sk(sk
)->dccps_pcrlen
;
792 if (dh
->dccph_cscov
&& (min_cov
== 0 || dh
->dccph_cscov
< min_cov
)) {
793 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
794 dh
->dccph_cscov
, min_cov
);
795 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
796 goto discard_and_relse
;
799 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
800 goto discard_and_relse
;
802 return sk_receive_skb(sk
, skb
, 1) ? -1 : 0;
805 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
810 * Generate Reset(No Connection) unless P.type == Reset
811 * Drop packet and return
813 if (dh
->dccph_type
!= DCCP_PKT_RESET
) {
814 DCCP_SKB_CB(skb
)->dccpd_reset_code
=
815 DCCP_RESET_CODE_NO_CONNECTION
;
816 dccp_v6_ctl_send_reset(sk
, skb
);
828 static int dccp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
831 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*)uaddr
;
832 struct inet_connection_sock
*icsk
= inet_csk(sk
);
833 struct inet_sock
*inet
= inet_sk(sk
);
834 struct ipv6_pinfo
*np
= inet6_sk(sk
);
835 struct dccp_sock
*dp
= dccp_sk(sk
);
836 struct in6_addr
*saddr
= NULL
, *final_p
, final
;
838 struct dst_entry
*dst
;
842 dp
->dccps_role
= DCCP_ROLE_CLIENT
;
844 if (addr_len
< SIN6_LEN_RFC2133
)
847 if (usin
->sin6_family
!= AF_INET6
)
848 return -EAFNOSUPPORT
;
850 memset(&fl6
, 0, sizeof(fl6
));
853 fl6
.flowlabel
= usin
->sin6_flowinfo
& IPV6_FLOWINFO_MASK
;
854 IP6_ECN_flow_init(fl6
.flowlabel
);
855 if (fl6
.flowlabel
& IPV6_FLOWLABEL_MASK
) {
856 struct ip6_flowlabel
*flowlabel
;
857 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
858 if (flowlabel
== NULL
)
860 fl6_sock_release(flowlabel
);
864 * connect() to INADDR_ANY means loopback (BSD'ism).
866 if (ipv6_addr_any(&usin
->sin6_addr
))
867 usin
->sin6_addr
.s6_addr
[15] = 1;
869 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
871 if (addr_type
& IPV6_ADDR_MULTICAST
)
874 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
875 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
876 usin
->sin6_scope_id
) {
877 /* If interface is set while binding, indices
880 if (sk
->sk_bound_dev_if
&&
881 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
884 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
887 /* Connect to link-local address requires an interface */
888 if (!sk
->sk_bound_dev_if
)
892 sk
->sk_v6_daddr
= usin
->sin6_addr
;
893 np
->flow_label
= fl6
.flowlabel
;
898 if (addr_type
== IPV6_ADDR_MAPPED
) {
899 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
900 struct sockaddr_in sin
;
902 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
904 if (__ipv6_only_sock(sk
))
907 sin
.sin_family
= AF_INET
;
908 sin
.sin_port
= usin
->sin6_port
;
909 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
911 icsk
->icsk_af_ops
= &dccp_ipv6_mapped
;
912 sk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
914 err
= dccp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
916 icsk
->icsk_ext_hdr_len
= exthdrlen
;
917 icsk
->icsk_af_ops
= &dccp_ipv6_af_ops
;
918 sk
->sk_backlog_rcv
= dccp_v6_do_rcv
;
921 ipv6_addr_set_v4mapped(inet
->inet_saddr
, &np
->saddr
);
922 ipv6_addr_set_v4mapped(inet
->inet_rcv_saddr
, &sk
->sk_v6_rcv_saddr
);
927 if (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
928 saddr
= &sk
->sk_v6_rcv_saddr
;
930 fl6
.flowi6_proto
= IPPROTO_DCCP
;
931 fl6
.daddr
= sk
->sk_v6_daddr
;
932 fl6
.saddr
= saddr
? *saddr
: np
->saddr
;
933 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
934 fl6
.fl6_dport
= usin
->sin6_port
;
935 fl6
.fl6_sport
= inet
->inet_sport
;
936 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
938 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
940 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
948 sk
->sk_v6_rcv_saddr
= *saddr
;
951 /* set the source address */
953 inet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
955 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
957 icsk
->icsk_ext_hdr_len
= 0;
959 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
962 inet
->inet_dport
= usin
->sin6_port
;
964 dccp_set_state(sk
, DCCP_REQUESTING
);
965 err
= inet6_hash_connect(&dccp_death_row
, sk
);
969 dp
->dccps_iss
= secure_dccpv6_sequence_number(np
->saddr
.s6_addr32
,
970 sk
->sk_v6_daddr
.s6_addr32
,
973 err
= dccp_connect(sk
);
980 dccp_set_state(sk
, DCCP_CLOSED
);
983 inet
->inet_dport
= 0;
984 sk
->sk_route_caps
= 0;
988 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops
= {
989 .queue_xmit
= inet6_csk_xmit
,
990 .send_check
= dccp_v6_send_check
,
991 .rebuild_header
= inet6_sk_rebuild_header
,
992 .conn_request
= dccp_v6_conn_request
,
993 .syn_recv_sock
= dccp_v6_request_recv_sock
,
994 .net_header_len
= sizeof(struct ipv6hdr
),
995 .setsockopt
= ipv6_setsockopt
,
996 .getsockopt
= ipv6_getsockopt
,
997 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
998 .sockaddr_len
= sizeof(struct sockaddr_in6
),
999 .bind_conflict
= inet6_csk_bind_conflict
,
1000 #ifdef CONFIG_COMPAT
1001 .compat_setsockopt
= compat_ipv6_setsockopt
,
1002 .compat_getsockopt
= compat_ipv6_getsockopt
,
1007 * DCCP over IPv4 via INET6 API
1009 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped
= {
1010 .queue_xmit
= ip_queue_xmit
,
1011 .send_check
= dccp_v4_send_check
,
1012 .rebuild_header
= inet_sk_rebuild_header
,
1013 .conn_request
= dccp_v6_conn_request
,
1014 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1015 .net_header_len
= sizeof(struct iphdr
),
1016 .setsockopt
= ipv6_setsockopt
,
1017 .getsockopt
= ipv6_getsockopt
,
1018 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1019 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1020 #ifdef CONFIG_COMPAT
1021 .compat_setsockopt
= compat_ipv6_setsockopt
,
1022 .compat_getsockopt
= compat_ipv6_getsockopt
,
1026 /* NOTE: A lot of things set to zero explicitly by call to
1027 * sk_alloc() so need not be done here.
1029 static int dccp_v6_init_sock(struct sock
*sk
)
1031 static __u8 dccp_v6_ctl_sock_initialized
;
1032 int err
= dccp_init_sock(sk
, dccp_v6_ctl_sock_initialized
);
1035 if (unlikely(!dccp_v6_ctl_sock_initialized
))
1036 dccp_v6_ctl_sock_initialized
= 1;
1037 inet_csk(sk
)->icsk_af_ops
= &dccp_ipv6_af_ops
;
1043 static void dccp_v6_destroy_sock(struct sock
*sk
)
1045 dccp_destroy_sock(sk
);
1046 inet6_destroy_sock(sk
);
1049 static struct timewait_sock_ops dccp6_timewait_sock_ops
= {
1050 .twsk_obj_size
= sizeof(struct dccp6_timewait_sock
),
1053 static struct proto dccp_v6_prot
= {
1055 .owner
= THIS_MODULE
,
1056 .close
= dccp_close
,
1057 .connect
= dccp_v6_connect
,
1058 .disconnect
= dccp_disconnect
,
1059 .ioctl
= dccp_ioctl
,
1060 .init
= dccp_v6_init_sock
,
1061 .setsockopt
= dccp_setsockopt
,
1062 .getsockopt
= dccp_getsockopt
,
1063 .sendmsg
= dccp_sendmsg
,
1064 .recvmsg
= dccp_recvmsg
,
1065 .backlog_rcv
= dccp_v6_do_rcv
,
1066 .hash
= dccp_v6_hash
,
1067 .unhash
= inet_unhash
,
1068 .accept
= inet_csk_accept
,
1069 .get_port
= inet_csk_get_port
,
1070 .shutdown
= dccp_shutdown
,
1071 .destroy
= dccp_v6_destroy_sock
,
1072 .orphan_count
= &dccp_orphan_count
,
1073 .max_header
= MAX_DCCP_HEADER
,
1074 .obj_size
= sizeof(struct dccp6_sock
),
1075 .slab_flags
= SLAB_DESTROY_BY_RCU
,
1076 .rsk_prot
= &dccp6_request_sock_ops
,
1077 .twsk_prot
= &dccp6_timewait_sock_ops
,
1078 .h
.hashinfo
= &dccp_hashinfo
,
1079 #ifdef CONFIG_COMPAT
1080 .compat_setsockopt
= compat_dccp_setsockopt
,
1081 .compat_getsockopt
= compat_dccp_getsockopt
,
1085 static const struct inet6_protocol dccp_v6_protocol
= {
1086 .handler
= dccp_v6_rcv
,
1087 .err_handler
= dccp_v6_err
,
1088 .flags
= INET6_PROTO_NOPOLICY
| INET6_PROTO_FINAL
,
1091 static const struct proto_ops inet6_dccp_ops
= {
1093 .owner
= THIS_MODULE
,
1094 .release
= inet6_release
,
1096 .connect
= inet_stream_connect
,
1097 .socketpair
= sock_no_socketpair
,
1098 .accept
= inet_accept
,
1099 .getname
= inet6_getname
,
1101 .ioctl
= inet6_ioctl
,
1102 .listen
= inet_dccp_listen
,
1103 .shutdown
= inet_shutdown
,
1104 .setsockopt
= sock_common_setsockopt
,
1105 .getsockopt
= sock_common_getsockopt
,
1106 .sendmsg
= inet_sendmsg
,
1107 .recvmsg
= sock_common_recvmsg
,
1108 .mmap
= sock_no_mmap
,
1109 .sendpage
= sock_no_sendpage
,
1110 #ifdef CONFIG_COMPAT
1111 .compat_setsockopt
= compat_sock_common_setsockopt
,
1112 .compat_getsockopt
= compat_sock_common_getsockopt
,
1116 static struct inet_protosw dccp_v6_protosw
= {
1118 .protocol
= IPPROTO_DCCP
,
1119 .prot
= &dccp_v6_prot
,
1120 .ops
= &inet6_dccp_ops
,
1121 .flags
= INET_PROTOSW_ICSK
,
1124 static int __net_init
dccp_v6_init_net(struct net
*net
)
1126 if (dccp_hashinfo
.bhash
== NULL
)
1127 return -ESOCKTNOSUPPORT
;
1129 return inet_ctl_sock_create(&net
->dccp
.v6_ctl_sk
, PF_INET6
,
1130 SOCK_DCCP
, IPPROTO_DCCP
, net
);
1133 static void __net_exit
dccp_v6_exit_net(struct net
*net
)
1135 inet_ctl_sock_destroy(net
->dccp
.v6_ctl_sk
);
1138 static struct pernet_operations dccp_v6_ops
= {
1139 .init
= dccp_v6_init_net
,
1140 .exit
= dccp_v6_exit_net
,
1143 static int __init
dccp_v6_init(void)
1145 int err
= proto_register(&dccp_v6_prot
, 1);
1150 err
= inet6_add_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1152 goto out_unregister_proto
;
1154 inet6_register_protosw(&dccp_v6_protosw
);
1156 err
= register_pernet_subsys(&dccp_v6_ops
);
1158 goto out_destroy_ctl_sock
;
1162 out_destroy_ctl_sock
:
1163 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1164 inet6_unregister_protosw(&dccp_v6_protosw
);
1165 out_unregister_proto
:
1166 proto_unregister(&dccp_v6_prot
);
1170 static void __exit
dccp_v6_exit(void)
1172 unregister_pernet_subsys(&dccp_v6_ops
);
1173 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1174 inet6_unregister_protosw(&dccp_v6_protosw
);
1175 proto_unregister(&dccp_v6_prot
);
1178 module_init(dccp_v6_init
);
1179 module_exit(dccp_v6_exit
);
1182 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1183 * values directly, Also cover the case where the protocol is not specified,
1184 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1186 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 33, 6);
1187 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6
, 0, 6);
1188 MODULE_LICENSE("GPL");
1189 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1190 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");