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 /* Socket used for sending RSTs and ACKs */
37 static struct socket
*dccp_v6_ctl_socket
;
39 static struct inet_connection_sock_af_ops dccp_ipv6_mapped
;
40 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops
;
42 static int dccp_v6_get_port(struct sock
*sk
, unsigned short snum
)
44 return inet_csk_get_port(&dccp_hashinfo
, sk
, snum
,
45 inet6_csk_bind_conflict
);
48 static void dccp_v6_hash(struct sock
*sk
)
50 if (sk
->sk_state
!= DCCP_CLOSED
) {
51 if (inet_csk(sk
)->icsk_af_ops
== &dccp_ipv6_mapped
) {
56 __inet6_hash(&dccp_hashinfo
, sk
);
61 /* add pseudo-header to DCCP checksum stored in skb->csum */
62 static inline u16
dccp_v6_csum_finish(struct sk_buff
*skb
,
63 struct in6_addr
*saddr
,
64 struct in6_addr
*daddr
)
66 return csum_ipv6_magic(saddr
, daddr
, skb
->len
, IPPROTO_DCCP
, skb
->csum
);
69 static inline void dccp_v6_send_check(struct sock
*sk
, int unused_value
,
72 struct ipv6_pinfo
*np
= inet6_sk(sk
);
73 struct dccp_hdr
*dh
= dccp_hdr(skb
);
75 dccp_csum_outgoing(skb
);
76 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
, &np
->saddr
, &np
->daddr
);
79 static inline __u32
secure_dccpv6_sequence_number(__u32
*saddr
, __u32
*daddr
,
80 __u16 sport
, __u16 dport
)
82 return secure_tcpv6_sequence_number(saddr
, daddr
, sport
, dport
);
85 static inline __u32
dccp_v6_init_sequence(struct sk_buff
*skb
)
87 return secure_dccpv6_sequence_number(skb
->nh
.ipv6h
->daddr
.s6_addr32
,
88 skb
->nh
.ipv6h
->saddr
.s6_addr32
,
89 dccp_hdr(skb
)->dccph_dport
,
90 dccp_hdr(skb
)->dccph_sport
);
94 static void dccp_v6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
95 int type
, int code
, int offset
, __be32 info
)
97 struct ipv6hdr
*hdr
= (struct ipv6hdr
*)skb
->data
;
98 const struct dccp_hdr
*dh
= (struct dccp_hdr
*)(skb
->data
+ offset
);
99 struct ipv6_pinfo
*np
;
104 sk
= inet6_lookup(&dccp_hashinfo
, &hdr
->daddr
, dh
->dccph_dport
,
105 &hdr
->saddr
, dh
->dccph_sport
, inet6_iif(skb
));
108 ICMP6_INC_STATS_BH(__in6_dev_get(skb
->dev
), ICMP6_MIB_INERRORS
);
112 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
113 inet_twsk_put(inet_twsk(sk
));
118 if (sock_owned_by_user(sk
))
119 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS
);
121 if (sk
->sk_state
== DCCP_CLOSED
)
126 if (type
== ICMPV6_PKT_TOOBIG
) {
127 struct dst_entry
*dst
= NULL
;
129 if (sock_owned_by_user(sk
))
131 if ((1 << sk
->sk_state
) & (DCCPF_LISTEN
| DCCPF_CLOSED
))
134 /* icmp should have updated the destination cache entry */
135 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
137 struct inet_sock
*inet
= inet_sk(sk
);
140 /* BUGGG_FUTURE: Again, it is not clear how
141 to handle rthdr case. Ignore this complexity
144 memset(&fl
, 0, sizeof(fl
));
145 fl
.proto
= IPPROTO_DCCP
;
146 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
147 ipv6_addr_copy(&fl
.fl6_src
, &np
->saddr
);
148 fl
.oif
= sk
->sk_bound_dev_if
;
149 fl
.fl_ip_dport
= inet
->dport
;
150 fl
.fl_ip_sport
= inet
->sport
;
151 security_sk_classify_flow(sk
, &fl
);
153 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
155 sk
->sk_err_soft
= -err
;
159 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
161 sk
->sk_err_soft
= -err
;
167 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
)) {
168 dccp_sync_mss(sk
, dst_mtu(dst
));
169 } /* else let the usual retransmit timer handle it */
174 icmpv6_err_convert(type
, code
, &err
);
176 seq
= DCCP_SKB_CB(skb
)->dccpd_seq
;
177 /* Might be for an request_sock */
178 switch (sk
->sk_state
) {
179 struct request_sock
*req
, **prev
;
181 if (sock_owned_by_user(sk
))
184 req
= inet6_csk_search_req(sk
, &prev
, dh
->dccph_dport
,
185 &hdr
->daddr
, &hdr
->saddr
,
191 * ICMPs are not backlogged, hence we cannot get an established
194 BUG_TRAP(req
->sk
== NULL
);
196 if (seq
!= dccp_rsk(req
)->dreq_iss
) {
197 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS
);
201 inet_csk_reqsk_queue_drop(sk
, req
, prev
);
204 case DCCP_REQUESTING
:
205 case DCCP_RESPOND
: /* Cannot happen.
206 It can, it SYNs are crossed. --ANK */
207 if (!sock_owned_by_user(sk
)) {
208 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
211 * Wake people up to see the error
212 * (see connect in sock.c)
214 sk
->sk_error_report(sk
);
217 sk
->sk_err_soft
= err
;
221 if (!sock_owned_by_user(sk
) && np
->recverr
) {
223 sk
->sk_error_report(sk
);
225 sk
->sk_err_soft
= err
;
233 static int dccp_v6_send_response(struct sock
*sk
, struct request_sock
*req
,
234 struct dst_entry
*dst
)
236 struct inet6_request_sock
*ireq6
= inet6_rsk(req
);
237 struct ipv6_pinfo
*np
= inet6_sk(sk
);
239 struct ipv6_txoptions
*opt
= NULL
;
240 struct in6_addr
*final_p
= NULL
, final
;
244 memset(&fl
, 0, sizeof(fl
));
245 fl
.proto
= IPPROTO_DCCP
;
246 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
247 ipv6_addr_copy(&fl
.fl6_src
, &ireq6
->loc_addr
);
248 fl
.fl6_flowlabel
= 0;
250 fl
.fl_ip_dport
= inet_rsk(req
)->rmt_port
;
251 fl
.fl_ip_sport
= inet_sk(sk
)->sport
;
252 security_req_classify_flow(req
, &fl
);
257 np
->rxopt
.bits
.osrcrt
== 2 &&
259 struct sk_buff
*pktopts
= ireq6
->pktopts
;
260 struct inet6_skb_parm
*rxopt
= IP6CB(pktopts
);
263 opt
= ipv6_invert_rthdr(sk
,
264 (struct ipv6_rt_hdr
*)(pktopts
->nh
.raw
+
268 if (opt
!= NULL
&& opt
->srcrt
!= NULL
) {
269 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)opt
->srcrt
;
271 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
272 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
276 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
281 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
283 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
288 skb
= dccp_make_response(sk
, dst
, req
);
290 struct dccp_hdr
*dh
= dccp_hdr(skb
);
292 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
,
295 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
296 err
= ip6_xmit(sk
, skb
, &fl
, opt
, 0);
297 if (err
== NET_XMIT_CN
)
302 if (opt
!= NULL
&& opt
!= np
->opt
)
303 sock_kfree_s(sk
, opt
, opt
->tot_len
);
308 static void dccp_v6_reqsk_destructor(struct request_sock
*req
)
310 if (inet6_rsk(req
)->pktopts
!= NULL
)
311 kfree_skb(inet6_rsk(req
)->pktopts
);
314 static void dccp_v6_ctl_send_reset(struct sk_buff
*rxskb
)
316 struct dccp_hdr
*rxdh
= dccp_hdr(rxskb
), *dh
;
317 const u32 dccp_hdr_reset_len
= sizeof(struct dccp_hdr
) +
318 sizeof(struct dccp_hdr_ext
) +
319 sizeof(struct dccp_hdr_reset
);
324 if (rxdh
->dccph_type
== DCCP_PKT_RESET
)
327 if (!ipv6_unicast_destination(rxskb
))
330 skb
= alloc_skb(dccp_v6_ctl_socket
->sk
->sk_prot
->max_header
,
335 skb_reserve(skb
, dccp_v6_ctl_socket
->sk
->sk_prot
->max_header
);
337 dh
= dccp_zeroed_hdr(skb
, dccp_hdr_reset_len
);
339 /* Swap the send and the receive. */
340 dh
->dccph_type
= DCCP_PKT_RESET
;
341 dh
->dccph_sport
= rxdh
->dccph_dport
;
342 dh
->dccph_dport
= rxdh
->dccph_sport
;
343 dh
->dccph_doff
= dccp_hdr_reset_len
/ 4;
345 dccp_hdr_reset(skb
)->dccph_reset_code
=
346 DCCP_SKB_CB(rxskb
)->dccpd_reset_code
;
348 /* See "8.3.1. Abnormal Termination" in RFC 4340 */
350 if (DCCP_SKB_CB(rxskb
)->dccpd_ack_seq
!= DCCP_PKT_WITHOUT_ACK_SEQ
)
351 dccp_set_seqno(&seqno
, DCCP_SKB_CB(rxskb
)->dccpd_ack_seq
+ 1);
353 dccp_hdr_set_seq(dh
, seqno
);
354 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb
),
355 DCCP_SKB_CB(rxskb
)->dccpd_seq
);
357 dccp_csum_outgoing(skb
);
358 dh
->dccph_checksum
= dccp_v6_csum_finish(skb
, &rxskb
->nh
.ipv6h
->saddr
,
359 &rxskb
->nh
.ipv6h
->daddr
);
361 memset(&fl
, 0, sizeof(fl
));
362 ipv6_addr_copy(&fl
.fl6_dst
, &rxskb
->nh
.ipv6h
->saddr
);
363 ipv6_addr_copy(&fl
.fl6_src
, &rxskb
->nh
.ipv6h
->daddr
);
365 fl
.proto
= IPPROTO_DCCP
;
366 fl
.oif
= inet6_iif(rxskb
);
367 fl
.fl_ip_dport
= dh
->dccph_dport
;
368 fl
.fl_ip_sport
= dh
->dccph_sport
;
369 security_skb_classify_flow(rxskb
, &fl
);
371 /* sk = NULL, but it is safe for now. RST socket required. */
372 if (!ip6_dst_lookup(NULL
, &skb
->dst
, &fl
)) {
373 if (xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) >= 0) {
374 ip6_xmit(dccp_v6_ctl_socket
->sk
, skb
, &fl
, NULL
, 0);
375 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS
);
376 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS
);
384 static struct request_sock_ops dccp6_request_sock_ops
= {
386 .obj_size
= sizeof(struct dccp6_request_sock
),
387 .rtx_syn_ack
= dccp_v6_send_response
,
388 .send_ack
= dccp_reqsk_send_ack
,
389 .destructor
= dccp_v6_reqsk_destructor
,
390 .send_reset
= dccp_v6_ctl_send_reset
,
393 static struct sock
*dccp_v6_hnd_req(struct sock
*sk
,struct sk_buff
*skb
)
395 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
396 const struct ipv6hdr
*iph
= skb
->nh
.ipv6h
;
398 struct request_sock
**prev
;
399 /* Find possible connection requests. */
400 struct request_sock
*req
= inet6_csk_search_req(sk
, &prev
,
406 return dccp_check_req(sk
, skb
, req
, prev
);
408 nsk
= __inet6_lookup_established(&dccp_hashinfo
,
409 &iph
->saddr
, dh
->dccph_sport
,
410 &iph
->daddr
, ntohs(dh
->dccph_dport
),
413 if (nsk
->sk_state
!= DCCP_TIME_WAIT
) {
417 inet_twsk_put(inet_twsk(nsk
));
424 static int dccp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
426 struct request_sock
*req
;
427 struct dccp_request_sock
*dreq
;
428 struct inet6_request_sock
*ireq6
;
429 struct ipv6_pinfo
*np
= inet6_sk(sk
);
430 const __be32 service
= dccp_hdr_request(skb
)->dccph_req_service
;
431 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
432 __u8 reset_code
= DCCP_RESET_CODE_TOO_BUSY
;
434 if (skb
->protocol
== htons(ETH_P_IP
))
435 return dccp_v4_conn_request(sk
, skb
);
437 if (!ipv6_unicast_destination(skb
))
440 if (dccp_bad_service_code(sk
, service
)) {
441 reset_code
= DCCP_RESET_CODE_BAD_SERVICE_CODE
;
445 * There are no SYN attacks on IPv6, yet...
447 if (inet_csk_reqsk_queue_is_full(sk
))
450 if (sk_acceptq_is_full(sk
) && inet_csk_reqsk_queue_young(sk
) > 1)
453 req
= inet6_reqsk_alloc(&dccp6_request_sock_ops
);
457 if (dccp_parse_options(sk
, skb
))
460 dccp_reqsk_init(req
, skb
);
462 if (security_inet_conn_request(sk
, skb
, req
))
465 ireq6
= inet6_rsk(req
);
466 ipv6_addr_copy(&ireq6
->rmt_addr
, &skb
->nh
.ipv6h
->saddr
);
467 ipv6_addr_copy(&ireq6
->loc_addr
, &skb
->nh
.ipv6h
->daddr
);
468 ireq6
->pktopts
= NULL
;
470 if (ipv6_opt_accepted(sk
, skb
) ||
471 np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
||
472 np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
) {
473 atomic_inc(&skb
->users
);
474 ireq6
->pktopts
= skb
;
476 ireq6
->iif
= sk
->sk_bound_dev_if
;
478 /* So that link locals have meaning */
479 if (!sk
->sk_bound_dev_if
&&
480 ipv6_addr_type(&ireq6
->rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
481 ireq6
->iif
= inet6_iif(skb
);
484 * Step 3: Process LISTEN state
486 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
488 * In fact we defer setting S.GSR, S.SWL, S.SWH to
489 * dccp_create_openreq_child.
491 dreq
= dccp_rsk(req
);
492 dreq
->dreq_isr
= dcb
->dccpd_seq
;
493 dreq
->dreq_iss
= dccp_v6_init_sequence(skb
);
494 dreq
->dreq_service
= service
;
496 if (dccp_v6_send_response(sk
, req
, NULL
))
499 inet6_csk_reqsk_queue_hash_add(sk
, req
, DCCP_TIMEOUT_INIT
);
505 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
506 dcb
->dccpd_reset_code
= reset_code
;
510 static struct sock
*dccp_v6_request_recv_sock(struct sock
*sk
,
512 struct request_sock
*req
,
513 struct dst_entry
*dst
)
515 struct inet6_request_sock
*ireq6
= inet6_rsk(req
);
516 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
517 struct inet_sock
*newinet
;
518 struct dccp_sock
*newdp
;
519 struct dccp6_sock
*newdp6
;
521 struct ipv6_txoptions
*opt
;
523 if (skb
->protocol
== htons(ETH_P_IP
)) {
527 newsk
= dccp_v4_request_recv_sock(sk
, skb
, req
, dst
);
531 newdp6
= (struct dccp6_sock
*)newsk
;
532 newdp
= dccp_sk(newsk
);
533 newinet
= inet_sk(newsk
);
534 newinet
->pinet6
= &newdp6
->inet6
;
535 newnp
= inet6_sk(newsk
);
537 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
539 ipv6_addr_set(&newnp
->daddr
, 0, 0, htonl(0x0000FFFF),
542 ipv6_addr_set(&newnp
->saddr
, 0, 0, htonl(0x0000FFFF),
545 ipv6_addr_copy(&newnp
->rcv_saddr
, &newnp
->saddr
);
547 inet_csk(newsk
)->icsk_af_ops
= &dccp_ipv6_mapped
;
548 newsk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
549 newnp
->pktoptions
= NULL
;
551 newnp
->mcast_oif
= inet6_iif(skb
);
552 newnp
->mcast_hops
= skb
->nh
.ipv6h
->hop_limit
;
555 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
556 * here, dccp_create_openreq_child now does this for us, see the comment in
557 * that function for the gory details. -acme
560 /* It is tricky place. Until this moment IPv4 tcp
561 worked with IPv6 icsk.icsk_af_ops.
564 dccp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
571 if (sk_acceptq_is_full(sk
))
574 if (np
->rxopt
.bits
.osrcrt
== 2 && opt
== NULL
&& ireq6
->pktopts
) {
575 const struct inet6_skb_parm
*rxopt
= IP6CB(ireq6
->pktopts
);
578 opt
= ipv6_invert_rthdr(sk
,
579 (struct ipv6_rt_hdr
*)(ireq6
->pktopts
->nh
.raw
+
584 struct in6_addr
*final_p
= NULL
, final
;
587 memset(&fl
, 0, sizeof(fl
));
588 fl
.proto
= IPPROTO_DCCP
;
589 ipv6_addr_copy(&fl
.fl6_dst
, &ireq6
->rmt_addr
);
590 if (opt
!= NULL
&& opt
->srcrt
!= NULL
) {
591 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)opt
->srcrt
;
593 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
594 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
597 ipv6_addr_copy(&fl
.fl6_src
, &ireq6
->loc_addr
);
598 fl
.oif
= sk
->sk_bound_dev_if
;
599 fl
.fl_ip_dport
= inet_rsk(req
)->rmt_port
;
600 fl
.fl_ip_sport
= inet_sk(sk
)->sport
;
601 security_sk_classify_flow(sk
, &fl
);
603 if (ip6_dst_lookup(sk
, &dst
, &fl
))
607 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
609 if ((xfrm_lookup(&dst
, &fl
, sk
, 0)) < 0)
613 newsk
= dccp_create_openreq_child(sk
, req
, skb
);
618 * No need to charge this sock to the relevant IPv6 refcnt debug socks
619 * count here, dccp_create_openreq_child now does this for us, see the
620 * comment in that function for the gory details. -acme
623 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
624 newsk
->sk_route_caps
= dst
->dev
->features
& ~(NETIF_F_IP_CSUM
|
626 newdp6
= (struct dccp6_sock
*)newsk
;
627 newinet
= inet_sk(newsk
);
628 newinet
->pinet6
= &newdp6
->inet6
;
629 newdp
= dccp_sk(newsk
);
630 newnp
= inet6_sk(newsk
);
632 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
634 ipv6_addr_copy(&newnp
->daddr
, &ireq6
->rmt_addr
);
635 ipv6_addr_copy(&newnp
->saddr
, &ireq6
->loc_addr
);
636 ipv6_addr_copy(&newnp
->rcv_saddr
, &ireq6
->loc_addr
);
637 newsk
->sk_bound_dev_if
= ireq6
->iif
;
639 /* Now IPv6 options...
641 First: no IPv4 options.
646 newnp
->rxopt
.all
= np
->rxopt
.all
;
648 /* Clone pktoptions received with SYN */
649 newnp
->pktoptions
= NULL
;
650 if (ireq6
->pktopts
!= NULL
) {
651 newnp
->pktoptions
= skb_clone(ireq6
->pktopts
, GFP_ATOMIC
);
652 kfree_skb(ireq6
->pktopts
);
653 ireq6
->pktopts
= NULL
;
654 if (newnp
->pktoptions
)
655 skb_set_owner_r(newnp
->pktoptions
, newsk
);
658 newnp
->mcast_oif
= inet6_iif(skb
);
659 newnp
->mcast_hops
= skb
->nh
.ipv6h
->hop_limit
;
662 * Clone native IPv6 options from listening socket (if any)
664 * Yes, keeping reference count would be much more clever, but we make
665 * one more one thing there: reattach optmem to newsk.
668 newnp
->opt
= ipv6_dup_options(newsk
, opt
);
670 sock_kfree_s(sk
, opt
, opt
->tot_len
);
673 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
674 if (newnp
->opt
!= NULL
)
675 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
676 newnp
->opt
->opt_flen
);
678 dccp_sync_mss(newsk
, dst_mtu(dst
));
680 newinet
->daddr
= newinet
->saddr
= newinet
->rcv_saddr
= LOOPBACK4_IPV6
;
682 __inet6_hash(&dccp_hashinfo
, newsk
);
683 inet_inherit_port(&dccp_hashinfo
, sk
, newsk
);
688 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS
);
690 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS
);
691 if (opt
!= NULL
&& opt
!= np
->opt
)
692 sock_kfree_s(sk
, opt
, opt
->tot_len
);
697 /* The socket must have it's spinlock held when we get
700 * We have a potential double-lock case here, so even when
701 * doing backlog processing we use the BH locking scheme.
702 * This is because we cannot sleep with the original spinlock
705 static int dccp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
707 struct ipv6_pinfo
*np
= inet6_sk(sk
);
708 struct sk_buff
*opt_skb
= NULL
;
710 /* Imagine: socket is IPv6. IPv4 packet arrives,
711 goes to IPv4 receive handler and backlogged.
712 From backlog it always goes here. Kerboom...
713 Fortunately, dccp_rcv_established and rcv_established
714 handle them correctly, but it is not case with
715 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
718 if (skb
->protocol
== htons(ETH_P_IP
))
719 return dccp_v4_do_rcv(sk
, skb
);
721 if (sk_filter(sk
, skb
))
725 * socket locking is here for SMP purposes as backlog rcv is currently
726 * called with bh processing disabled.
729 /* Do Stevens' IPV6_PKTOPTIONS.
731 Yes, guys, it is the only place in our code, where we
732 may make it not affecting IPv4.
733 The rest of code is protocol independent,
734 and I do not like idea to uglify IPv4.
736 Actually, all the idea behind IPV6_PKTOPTIONS
737 looks not very well thought. For now we latch
738 options, received in the last packet, enqueued
739 by tcp. Feel free to propose better solution.
744 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
745 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
747 opt_skb
= skb_clone(skb
, GFP_ATOMIC
);
749 if (sk
->sk_state
== DCCP_OPEN
) { /* Fast path */
750 if (dccp_rcv_established(sk
, skb
, dccp_hdr(skb
), skb
->len
))
753 /* XXX This is where we would goto ipv6_pktoptions. */
754 __kfree_skb(opt_skb
);
760 * Step 3: Process LISTEN state
761 * If S.state == LISTEN,
762 * If P.type == Request or P contains a valid Init Cookie option,
763 * (* Must scan the packet's options to check for Init
764 * Cookies. Only Init Cookies are processed here,
765 * however; other options are processed in Step 8. This
766 * scan need only be performed if the endpoint uses Init
768 * (* Generate a new socket and switch to that socket *)
769 * Set S := new socket for this port pair
771 * Choose S.ISS (initial seqno) or set from Init Cookies
772 * Initialize S.GAR := S.ISS
773 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
774 * Continue with S.state == RESPOND
775 * (* A Response packet will be generated in Step 11 *)
777 * Generate Reset(No Connection) unless P.type == Reset
778 * Drop packet and return
780 * NOTE: the check for the packet types is done in
781 * dccp_rcv_state_process
783 if (sk
->sk_state
== DCCP_LISTEN
) {
784 struct sock
*nsk
= dccp_v6_hnd_req(sk
, skb
);
789 * Queue it on the new socket if the new socket is active,
790 * otherwise we just shortcircuit this and continue with
794 if (dccp_child_process(sk
, nsk
, skb
))
797 __kfree_skb(opt_skb
);
802 if (dccp_rcv_state_process(sk
, skb
, dccp_hdr(skb
), skb
->len
))
805 /* XXX This is where we would goto ipv6_pktoptions. */
806 __kfree_skb(opt_skb
);
811 dccp_v6_ctl_send_reset(skb
);
814 __kfree_skb(opt_skb
);
819 static int dccp_v6_rcv(struct sk_buff
**pskb
)
821 const struct dccp_hdr
*dh
;
822 struct sk_buff
*skb
= *pskb
;
826 /* Step 1: Check header basics */
828 if (dccp_invalid_packet(skb
))
831 /* Step 1: If header checksum is incorrect, drop packet and return. */
832 if (dccp_v6_csum_finish(skb
, &skb
->nh
.ipv6h
->saddr
,
833 &skb
->nh
.ipv6h
->daddr
)) {
834 LIMIT_NETDEBUG(KERN_WARNING
835 "%s: dropped packet with invalid checksum\n",
842 DCCP_SKB_CB(skb
)->dccpd_seq
= dccp_hdr_seq(skb
);
843 DCCP_SKB_CB(skb
)->dccpd_type
= dh
->dccph_type
;
845 if (dccp_packet_without_ack(skb
))
846 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= DCCP_PKT_WITHOUT_ACK_SEQ
;
848 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= dccp_hdr_ack_seq(skb
);
851 * Look up flow ID in table and get corresponding socket */
852 sk
= __inet6_lookup(&dccp_hashinfo
, &skb
->nh
.ipv6h
->saddr
,
854 &skb
->nh
.ipv6h
->daddr
, ntohs(dh
->dccph_dport
),
861 dccp_pr_debug("failed to look up flow ID in table and "
862 "get corresponding socket\n");
868 * ... or S.state == TIMEWAIT,
869 * Generate Reset(No Connection) unless P.type == Reset
870 * Drop packet and return
872 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
873 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
874 inet_twsk_put(inet_twsk(sk
));
879 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
880 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
881 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
883 min_cov
= dccp_sk(sk
)->dccps_pcrlen
;
884 if (dh
->dccph_cscov
&& (min_cov
== 0 || dh
->dccph_cscov
< min_cov
)) {
885 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
886 dh
->dccph_cscov
, min_cov
);
887 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
888 goto discard_and_relse
;
891 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
892 goto discard_and_relse
;
894 return sk_receive_skb(sk
, skb
) ? -1 : 0;
897 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
902 * Generate Reset(No Connection) unless P.type == Reset
903 * Drop packet and return
905 if (dh
->dccph_type
!= DCCP_PKT_RESET
) {
906 DCCP_SKB_CB(skb
)->dccpd_reset_code
=
907 DCCP_RESET_CODE_NO_CONNECTION
;
908 dccp_v6_ctl_send_reset(skb
);
920 static int dccp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
923 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*)uaddr
;
924 struct inet_connection_sock
*icsk
= inet_csk(sk
);
925 struct inet_sock
*inet
= inet_sk(sk
);
926 struct ipv6_pinfo
*np
= inet6_sk(sk
);
927 struct dccp_sock
*dp
= dccp_sk(sk
);
928 struct in6_addr
*saddr
= NULL
, *final_p
= NULL
, final
;
930 struct dst_entry
*dst
;
934 dp
->dccps_role
= DCCP_ROLE_CLIENT
;
936 if (addr_len
< SIN6_LEN_RFC2133
)
939 if (usin
->sin6_family
!= AF_INET6
)
940 return -EAFNOSUPPORT
;
942 memset(&fl
, 0, sizeof(fl
));
945 fl
.fl6_flowlabel
= usin
->sin6_flowinfo
& IPV6_FLOWINFO_MASK
;
946 IP6_ECN_flow_init(fl
.fl6_flowlabel
);
947 if (fl
.fl6_flowlabel
& IPV6_FLOWLABEL_MASK
) {
948 struct ip6_flowlabel
*flowlabel
;
949 flowlabel
= fl6_sock_lookup(sk
, fl
.fl6_flowlabel
);
950 if (flowlabel
== NULL
)
952 ipv6_addr_copy(&usin
->sin6_addr
, &flowlabel
->dst
);
953 fl6_sock_release(flowlabel
);
957 * connect() to INADDR_ANY means loopback (BSD'ism).
959 if (ipv6_addr_any(&usin
->sin6_addr
))
960 usin
->sin6_addr
.s6_addr
[15] = 1;
962 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
964 if (addr_type
& IPV6_ADDR_MULTICAST
)
967 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
968 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
969 usin
->sin6_scope_id
) {
970 /* If interface is set while binding, indices
973 if (sk
->sk_bound_dev_if
&&
974 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
977 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
980 /* Connect to link-local address requires an interface */
981 if (!sk
->sk_bound_dev_if
)
985 ipv6_addr_copy(&np
->daddr
, &usin
->sin6_addr
);
986 np
->flow_label
= fl
.fl6_flowlabel
;
991 if (addr_type
== IPV6_ADDR_MAPPED
) {
992 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
993 struct sockaddr_in sin
;
995 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
997 if (__ipv6_only_sock(sk
))
1000 sin
.sin_family
= AF_INET
;
1001 sin
.sin_port
= usin
->sin6_port
;
1002 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
1004 icsk
->icsk_af_ops
= &dccp_ipv6_mapped
;
1005 sk
->sk_backlog_rcv
= dccp_v4_do_rcv
;
1007 err
= dccp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
1009 icsk
->icsk_ext_hdr_len
= exthdrlen
;
1010 icsk
->icsk_af_ops
= &dccp_ipv6_af_ops
;
1011 sk
->sk_backlog_rcv
= dccp_v6_do_rcv
;
1014 ipv6_addr_set(&np
->saddr
, 0, 0, htonl(0x0000FFFF),
1016 ipv6_addr_set(&np
->rcv_saddr
, 0, 0, htonl(0x0000FFFF),
1023 if (!ipv6_addr_any(&np
->rcv_saddr
))
1024 saddr
= &np
->rcv_saddr
;
1026 fl
.proto
= IPPROTO_DCCP
;
1027 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
1028 ipv6_addr_copy(&fl
.fl6_src
, saddr
? saddr
: &np
->saddr
);
1029 fl
.oif
= sk
->sk_bound_dev_if
;
1030 fl
.fl_ip_dport
= usin
->sin6_port
;
1031 fl
.fl_ip_sport
= inet
->sport
;
1032 security_sk_classify_flow(sk
, &fl
);
1034 if (np
->opt
!= NULL
&& np
->opt
->srcrt
!= NULL
) {
1035 const struct rt0_hdr
*rt0
= (struct rt0_hdr
*)np
->opt
->srcrt
;
1037 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
1038 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
1042 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
1047 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
1049 err
= xfrm_lookup(&dst
, &fl
, sk
, 0);
1053 if (saddr
== NULL
) {
1054 saddr
= &fl
.fl6_src
;
1055 ipv6_addr_copy(&np
->rcv_saddr
, saddr
);
1058 /* set the source address */
1059 ipv6_addr_copy(&np
->saddr
, saddr
);
1060 inet
->rcv_saddr
= LOOPBACK4_IPV6
;
1062 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
1064 icsk
->icsk_ext_hdr_len
= 0;
1065 if (np
->opt
!= NULL
)
1066 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
1067 np
->opt
->opt_nflen
);
1069 inet
->dport
= usin
->sin6_port
;
1071 dccp_set_state(sk
, DCCP_REQUESTING
);
1072 err
= inet6_hash_connect(&dccp_death_row
, sk
);
1076 dp
->dccps_iss
= secure_dccpv6_sequence_number(np
->saddr
.s6_addr32
,
1077 np
->daddr
.s6_addr32
,
1078 inet
->sport
, inet
->dport
);
1079 err
= dccp_connect(sk
);
1086 dccp_set_state(sk
, DCCP_CLOSED
);
1090 sk
->sk_route_caps
= 0;
1094 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops
= {
1095 .queue_xmit
= inet6_csk_xmit
,
1096 .send_check
= dccp_v6_send_check
,
1097 .rebuild_header
= inet6_sk_rebuild_header
,
1098 .conn_request
= dccp_v6_conn_request
,
1099 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1100 .net_header_len
= sizeof(struct ipv6hdr
),
1101 .setsockopt
= ipv6_setsockopt
,
1102 .getsockopt
= ipv6_getsockopt
,
1103 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1104 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1105 #ifdef CONFIG_COMPAT
1106 .compat_setsockopt
= compat_ipv6_setsockopt
,
1107 .compat_getsockopt
= compat_ipv6_getsockopt
,
1112 * DCCP over IPv4 via INET6 API
1114 static struct inet_connection_sock_af_ops dccp_ipv6_mapped
= {
1115 .queue_xmit
= ip_queue_xmit
,
1116 .send_check
= dccp_v4_send_check
,
1117 .rebuild_header
= inet_sk_rebuild_header
,
1118 .conn_request
= dccp_v6_conn_request
,
1119 .syn_recv_sock
= dccp_v6_request_recv_sock
,
1120 .net_header_len
= sizeof(struct iphdr
),
1121 .setsockopt
= ipv6_setsockopt
,
1122 .getsockopt
= ipv6_getsockopt
,
1123 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1124 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1125 #ifdef CONFIG_COMPAT
1126 .compat_setsockopt
= compat_ipv6_setsockopt
,
1127 .compat_getsockopt
= compat_ipv6_getsockopt
,
1131 /* NOTE: A lot of things set to zero explicitly by call to
1132 * sk_alloc() so need not be done here.
1134 static int dccp_v6_init_sock(struct sock
*sk
)
1136 static __u8 dccp_v6_ctl_sock_initialized
;
1137 int err
= dccp_init_sock(sk
, dccp_v6_ctl_sock_initialized
);
1140 if (unlikely(!dccp_v6_ctl_sock_initialized
))
1141 dccp_v6_ctl_sock_initialized
= 1;
1142 inet_csk(sk
)->icsk_af_ops
= &dccp_ipv6_af_ops
;
1148 static int dccp_v6_destroy_sock(struct sock
*sk
)
1150 dccp_destroy_sock(sk
);
1151 return inet6_destroy_sock(sk
);
1154 static struct timewait_sock_ops dccp6_timewait_sock_ops
= {
1155 .twsk_obj_size
= sizeof(struct dccp6_timewait_sock
),
1158 static struct proto dccp_v6_prot
= {
1160 .owner
= THIS_MODULE
,
1161 .close
= dccp_close
,
1162 .connect
= dccp_v6_connect
,
1163 .disconnect
= dccp_disconnect
,
1164 .ioctl
= dccp_ioctl
,
1165 .init
= dccp_v6_init_sock
,
1166 .setsockopt
= dccp_setsockopt
,
1167 .getsockopt
= dccp_getsockopt
,
1168 .sendmsg
= dccp_sendmsg
,
1169 .recvmsg
= dccp_recvmsg
,
1170 .backlog_rcv
= dccp_v6_do_rcv
,
1171 .hash
= dccp_v6_hash
,
1172 .unhash
= dccp_unhash
,
1173 .accept
= inet_csk_accept
,
1174 .get_port
= dccp_v6_get_port
,
1175 .shutdown
= dccp_shutdown
,
1176 .destroy
= dccp_v6_destroy_sock
,
1177 .orphan_count
= &dccp_orphan_count
,
1178 .max_header
= MAX_DCCP_HEADER
,
1179 .obj_size
= sizeof(struct dccp6_sock
),
1180 .rsk_prot
= &dccp6_request_sock_ops
,
1181 .twsk_prot
= &dccp6_timewait_sock_ops
,
1182 #ifdef CONFIG_COMPAT
1183 .compat_setsockopt
= compat_dccp_setsockopt
,
1184 .compat_getsockopt
= compat_dccp_getsockopt
,
1188 static struct inet6_protocol dccp_v6_protocol
= {
1189 .handler
= dccp_v6_rcv
,
1190 .err_handler
= dccp_v6_err
,
1191 .flags
= INET6_PROTO_NOPOLICY
| INET6_PROTO_FINAL
,
1194 static struct proto_ops inet6_dccp_ops
= {
1196 .owner
= THIS_MODULE
,
1197 .release
= inet6_release
,
1199 .connect
= inet_stream_connect
,
1200 .socketpair
= sock_no_socketpair
,
1201 .accept
= inet_accept
,
1202 .getname
= inet6_getname
,
1204 .ioctl
= inet6_ioctl
,
1205 .listen
= inet_dccp_listen
,
1206 .shutdown
= inet_shutdown
,
1207 .setsockopt
= sock_common_setsockopt
,
1208 .getsockopt
= sock_common_getsockopt
,
1209 .sendmsg
= inet_sendmsg
,
1210 .recvmsg
= sock_common_recvmsg
,
1211 .mmap
= sock_no_mmap
,
1212 .sendpage
= sock_no_sendpage
,
1213 #ifdef CONFIG_COMPAT
1214 .compat_setsockopt
= compat_sock_common_setsockopt
,
1215 .compat_getsockopt
= compat_sock_common_getsockopt
,
1219 static struct inet_protosw dccp_v6_protosw
= {
1221 .protocol
= IPPROTO_DCCP
,
1222 .prot
= &dccp_v6_prot
,
1223 .ops
= &inet6_dccp_ops
,
1225 .flags
= INET_PROTOSW_ICSK
,
1228 static int __init
dccp_v6_init(void)
1230 int err
= proto_register(&dccp_v6_prot
, 1);
1235 err
= inet6_add_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1237 goto out_unregister_proto
;
1239 inet6_register_protosw(&dccp_v6_protosw
);
1241 err
= inet_csk_ctl_sock_create(&dccp_v6_ctl_socket
, PF_INET6
,
1242 SOCK_DCCP
, IPPROTO_DCCP
);
1244 goto out_unregister_protosw
;
1247 out_unregister_protosw
:
1248 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1249 inet6_unregister_protosw(&dccp_v6_protosw
);
1250 out_unregister_proto
:
1251 proto_unregister(&dccp_v6_prot
);
1255 static void __exit
dccp_v6_exit(void)
1257 inet6_del_protocol(&dccp_v6_protocol
, IPPROTO_DCCP
);
1258 inet6_unregister_protosw(&dccp_v6_protosw
);
1259 proto_unregister(&dccp_v6_prot
);
1262 module_init(dccp_v6_init
);
1263 module_exit(dccp_v6_exit
);
1266 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1267 * values directly, Also cover the case where the protocol is not specified,
1268 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1270 MODULE_ALIAS("net-pf-" __stringify(PF_INET6
) "-proto-33-type-6");
1271 MODULE_ALIAS("net-pf-" __stringify(PF_INET6
) "-proto-0-type-6");
1272 MODULE_LICENSE("GPL");
1273 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1274 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");