4 * An implementation of the DCCP protocol
5 * Arnaldo Carvalho de Melo <acme@conectiva.com.br>
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
13 #include <linux/dccp.h>
14 #include <linux/icmp.h>
15 #include <linux/slab.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/random.h>
21 #include <net/inet_common.h>
22 #include <net/inet_hashtables.h>
23 #include <net/inet_sock.h>
24 #include <net/protocol.h>
26 #include <net/timewait_sock.h>
27 #include <net/tcp_states.h>
29 #include <net/secure_seq.h>
37 * The per-net dccp.v4_ctl_sk socket is used for responding to
38 * the Out-of-the-blue (OOTB) packets. A control sock will be created
39 * for this socket at the initialization time.
42 int dccp_v4_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
44 const struct sockaddr_in
*usin
= (struct sockaddr_in
*)uaddr
;
45 struct inet_sock
*inet
= inet_sk(sk
);
46 struct dccp_sock
*dp
= dccp_sk(sk
);
47 __be16 orig_sport
, orig_dport
;
48 __be32 daddr
, nexthop
;
52 struct ip_options_rcu
*inet_opt
;
54 dp
->dccps_role
= DCCP_ROLE_CLIENT
;
56 if (addr_len
< sizeof(struct sockaddr_in
))
59 if (usin
->sin_family
!= AF_INET
)
62 nexthop
= daddr
= usin
->sin_addr
.s_addr
;
64 inet_opt
= rcu_dereference_protected(inet
->inet_opt
,
65 sock_owned_by_user(sk
));
66 if (inet_opt
!= NULL
&& inet_opt
->opt
.srr
) {
69 nexthop
= inet_opt
->opt
.faddr
;
72 orig_sport
= inet
->inet_sport
;
73 orig_dport
= usin
->sin_port
;
74 fl4
= &inet
->cork
.fl
.u
.ip4
;
75 rt
= ip_route_connect(fl4
, nexthop
, inet
->inet_saddr
,
76 RT_CONN_FLAGS(sk
), sk
->sk_bound_dev_if
,
78 orig_sport
, orig_dport
, sk
);
82 if (rt
->rt_flags
& (RTCF_MULTICAST
| RTCF_BROADCAST
)) {
87 if (inet_opt
== NULL
|| !inet_opt
->opt
.srr
)
90 if (inet
->inet_saddr
== 0)
91 inet
->inet_saddr
= fl4
->saddr
;
92 sk_rcv_saddr_set(sk
, inet
->inet_saddr
);
93 inet
->inet_dport
= usin
->sin_port
;
94 sk_daddr_set(sk
, daddr
);
96 inet_csk(sk
)->icsk_ext_hdr_len
= 0;
98 inet_csk(sk
)->icsk_ext_hdr_len
= inet_opt
->opt
.optlen
;
100 * Socket identity is still unknown (sport may be zero).
101 * However we set state to DCCP_REQUESTING and not releasing socket
102 * lock select source port, enter ourselves into the hash tables and
103 * complete initialization after this.
105 dccp_set_state(sk
, DCCP_REQUESTING
);
106 err
= inet_hash_connect(&dccp_death_row
, sk
);
110 rt
= ip_route_newports(fl4
, rt
, orig_sport
, orig_dport
,
111 inet
->inet_sport
, inet
->inet_dport
, sk
);
117 /* OK, now commit destination to socket. */
118 sk_setup_caps(sk
, &rt
->dst
);
120 dp
->dccps_iss
= secure_dccp_sequence_number(inet
->inet_saddr
,
124 inet
->inet_id
= dp
->dccps_iss
^ jiffies
;
126 err
= dccp_connect(sk
);
134 * This unhashes the socket and releases the local port, if necessary.
136 dccp_set_state(sk
, DCCP_CLOSED
);
138 sk
->sk_route_caps
= 0;
139 inet
->inet_dport
= 0;
142 EXPORT_SYMBOL_GPL(dccp_v4_connect
);
145 * This routine does path mtu discovery as defined in RFC1191.
147 static inline void dccp_do_pmtu_discovery(struct sock
*sk
,
148 const struct iphdr
*iph
,
151 struct dst_entry
*dst
;
152 const struct inet_sock
*inet
= inet_sk(sk
);
153 const struct dccp_sock
*dp
= dccp_sk(sk
);
155 /* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
156 * send out by Linux are always < 576bytes so they should go through
159 if (sk
->sk_state
== DCCP_LISTEN
)
162 dst
= inet_csk_update_pmtu(sk
, mtu
);
166 /* Something is about to be wrong... Remember soft error
167 * for the case, if this connection will not able to recover.
169 if (mtu
< dst_mtu(dst
) && ip_dont_fragment(sk
, dst
))
170 sk
->sk_err_soft
= EMSGSIZE
;
174 if (inet
->pmtudisc
!= IP_PMTUDISC_DONT
&&
175 ip_sk_accept_pmtu(sk
) &&
176 inet_csk(sk
)->icsk_pmtu_cookie
> mtu
) {
177 dccp_sync_mss(sk
, mtu
);
180 * From RFC 4340, sec. 14.1:
182 * DCCP-Sync packets are the best choice for upward
183 * probing, since DCCP-Sync probes do not risk application
186 dccp_send_sync(sk
, dp
->dccps_gsr
, DCCP_PKT_SYNC
);
187 } /* else let the usual retransmit timer handle it */
190 static void dccp_do_redirect(struct sk_buff
*skb
, struct sock
*sk
)
192 struct dst_entry
*dst
= __sk_dst_check(sk
, 0);
195 dst
->ops
->redirect(dst
, sk
, skb
);
199 * This routine is called by the ICMP module when it gets some sort of error
200 * condition. If err < 0 then the socket should be closed and the error
201 * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
202 * After adjustment header points to the first 8 bytes of the tcp header. We
203 * need to find the appropriate port.
205 * The locking strategy used here is very "optimistic". When someone else
206 * accesses the socket the ICMP is just dropped and for some paths there is no
207 * check at all. A more general error queue to queue errors for later handling
208 * is probably better.
210 static void dccp_v4_err(struct sk_buff
*skb
, u32 info
)
212 const struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
213 const u8 offset
= iph
->ihl
<< 2;
214 const struct dccp_hdr
*dh
= (struct dccp_hdr
*)(skb
->data
+ offset
);
215 struct dccp_sock
*dp
;
216 struct inet_sock
*inet
;
217 const int type
= icmp_hdr(skb
)->type
;
218 const int code
= icmp_hdr(skb
)->code
;
222 struct net
*net
= dev_net(skb
->dev
);
224 if (skb
->len
< offset
+ sizeof(*dh
) ||
225 skb
->len
< offset
+ __dccp_basic_hdr_len(dh
)) {
226 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
230 sk
= inet_lookup(net
, &dccp_hashinfo
,
231 iph
->daddr
, dh
->dccph_dport
,
232 iph
->saddr
, dh
->dccph_sport
, inet_iif(skb
));
234 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
238 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
239 inet_twsk_put(inet_twsk(sk
));
244 /* If too many ICMPs get dropped on busy
245 * servers this needs to be solved differently.
247 if (sock_owned_by_user(sk
))
248 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
250 if (sk
->sk_state
== DCCP_CLOSED
)
254 seq
= dccp_hdr_seq(dh
);
255 if ((1 << sk
->sk_state
) & ~(DCCPF_REQUESTING
| DCCPF_LISTEN
) &&
256 !between48(seq
, dp
->dccps_awl
, dp
->dccps_awh
)) {
257 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
263 dccp_do_redirect(skb
, sk
);
265 case ICMP_SOURCE_QUENCH
:
266 /* Just silently ignore these. */
268 case ICMP_PARAMETERPROB
:
271 case ICMP_DEST_UNREACH
:
272 if (code
> NR_ICMP_UNREACH
)
275 if (code
== ICMP_FRAG_NEEDED
) { /* PMTU discovery (RFC1191) */
276 if (!sock_owned_by_user(sk
))
277 dccp_do_pmtu_discovery(sk
, iph
, info
);
281 err
= icmp_err_convert
[code
].errno
;
283 case ICMP_TIME_EXCEEDED
:
290 switch (sk
->sk_state
) {
291 struct request_sock
*req
;
293 if (sock_owned_by_user(sk
))
295 req
= inet_csk_search_req(sk
, dh
->dccph_dport
,
296 iph
->daddr
, iph
->saddr
);
301 * ICMPs are not backlogged, hence we cannot get an established
306 if (!between48(seq
, dccp_rsk(req
)->dreq_iss
,
307 dccp_rsk(req
)->dreq_gss
)) {
308 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
312 * Still in RESPOND, just remove it silently.
313 * There is no good way to pass the error to the newly
314 * created socket, and POSIX does not want network
315 * errors returned from accept().
317 inet_csk_reqsk_queue_drop(sk
, req
);
320 case DCCP_REQUESTING
:
322 if (!sock_owned_by_user(sk
)) {
323 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
326 sk
->sk_error_report(sk
);
330 sk
->sk_err_soft
= err
;
334 /* If we've already connected we will keep trying
335 * until we time out, or the user gives up.
337 * rfc1122 4.2.3.9 allows to consider as hard errors
338 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
339 * but it is obsoleted by pmtu discovery).
341 * Note, that in modern internet, where routing is unreliable
342 * and in each dark corner broken firewalls sit, sending random
343 * errors ordered by their masters even this two messages finally lose
344 * their original sense (even Linux sends invalid PORT_UNREACHs)
346 * Now we are in compliance with RFCs.
351 if (!sock_owned_by_user(sk
) && inet
->recverr
) {
353 sk
->sk_error_report(sk
);
354 } else /* Only an error on timeout */
355 sk
->sk_err_soft
= err
;
361 static inline __sum16
dccp_v4_csum_finish(struct sk_buff
*skb
,
362 __be32 src
, __be32 dst
)
364 return csum_tcpudp_magic(src
, dst
, skb
->len
, IPPROTO_DCCP
, skb
->csum
);
367 void dccp_v4_send_check(struct sock
*sk
, struct sk_buff
*skb
)
369 const struct inet_sock
*inet
= inet_sk(sk
);
370 struct dccp_hdr
*dh
= dccp_hdr(skb
);
372 dccp_csum_outgoing(skb
);
373 dh
->dccph_checksum
= dccp_v4_csum_finish(skb
,
377 EXPORT_SYMBOL_GPL(dccp_v4_send_check
);
379 static inline u64
dccp_v4_init_sequence(const struct sk_buff
*skb
)
381 return secure_dccp_sequence_number(ip_hdr(skb
)->daddr
,
383 dccp_hdr(skb
)->dccph_dport
,
384 dccp_hdr(skb
)->dccph_sport
);
388 * The three way handshake has completed - we got a valid ACK or DATAACK -
389 * now create the new socket.
391 * This is the equivalent of TCP's tcp_v4_syn_recv_sock
393 struct sock
*dccp_v4_request_recv_sock(struct sock
*sk
, struct sk_buff
*skb
,
394 struct request_sock
*req
,
395 struct dst_entry
*dst
)
397 struct inet_request_sock
*ireq
;
398 struct inet_sock
*newinet
;
401 if (sk_acceptq_is_full(sk
))
404 newsk
= dccp_create_openreq_child(sk
, req
, skb
);
408 newinet
= inet_sk(newsk
);
409 ireq
= inet_rsk(req
);
410 sk_daddr_set(newsk
, ireq
->ir_rmt_addr
);
411 sk_rcv_saddr_set(newsk
, ireq
->ir_loc_addr
);
412 newinet
->inet_saddr
= ireq
->ir_loc_addr
;
413 newinet
->inet_opt
= ireq
->opt
;
415 newinet
->mc_index
= inet_iif(skb
);
416 newinet
->mc_ttl
= ip_hdr(skb
)->ttl
;
417 newinet
->inet_id
= jiffies
;
419 if (dst
== NULL
&& (dst
= inet_csk_route_child_sock(sk
, newsk
, req
)) == NULL
)
422 sk_setup_caps(newsk
, dst
);
424 dccp_sync_mss(newsk
, dst_mtu(dst
));
426 if (__inet_inherit_port(sk
, newsk
) < 0)
428 __inet_hash_nolisten(newsk
, NULL
);
433 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
437 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
440 inet_csk_prepare_forced_close(newsk
);
444 EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock
);
446 static struct sock
*dccp_v4_hnd_req(struct sock
*sk
, struct sk_buff
*skb
)
448 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
449 const struct iphdr
*iph
= ip_hdr(skb
);
451 /* Find possible connection requests. */
452 struct request_sock
*req
= inet_csk_search_req(sk
, dh
->dccph_sport
,
453 iph
->saddr
, iph
->daddr
);
455 return dccp_check_req(sk
, skb
, req
);
457 nsk
= inet_lookup_established(sock_net(sk
), &dccp_hashinfo
,
458 iph
->saddr
, dh
->dccph_sport
,
459 iph
->daddr
, dh
->dccph_dport
,
462 if (nsk
->sk_state
!= DCCP_TIME_WAIT
) {
466 inet_twsk_put(inet_twsk(nsk
));
473 static struct dst_entry
* dccp_v4_route_skb(struct net
*net
, struct sock
*sk
,
477 const struct iphdr
*iph
= ip_hdr(skb
);
478 struct flowi4 fl4
= {
479 .flowi4_oif
= inet_iif(skb
),
482 .flowi4_tos
= RT_CONN_FLAGS(sk
),
483 .flowi4_proto
= sk
->sk_protocol
,
484 .fl4_sport
= dccp_hdr(skb
)->dccph_dport
,
485 .fl4_dport
= dccp_hdr(skb
)->dccph_sport
,
488 security_skb_classify_flow(skb
, flowi4_to_flowi(&fl4
));
489 rt
= ip_route_output_flow(net
, &fl4
, sk
);
491 IP_INC_STATS_BH(net
, IPSTATS_MIB_OUTNOROUTES
);
498 static int dccp_v4_send_response(struct sock
*sk
, struct request_sock
*req
)
502 struct dst_entry
*dst
;
505 dst
= inet_csk_route_req(sk
, &fl4
, req
);
509 skb
= dccp_make_response(sk
, dst
, req
);
511 const struct inet_request_sock
*ireq
= inet_rsk(req
);
512 struct dccp_hdr
*dh
= dccp_hdr(skb
);
514 dh
->dccph_checksum
= dccp_v4_csum_finish(skb
, ireq
->ir_loc_addr
,
516 err
= ip_build_and_send_pkt(skb
, sk
, ireq
->ir_loc_addr
,
519 err
= net_xmit_eval(err
);
527 static void dccp_v4_ctl_send_reset(struct sock
*sk
, struct sk_buff
*rxskb
)
530 const struct iphdr
*rxiph
;
532 struct dst_entry
*dst
;
533 struct net
*net
= dev_net(skb_dst(rxskb
)->dev
);
534 struct sock
*ctl_sk
= net
->dccp
.v4_ctl_sk
;
536 /* Never send a reset in response to a reset. */
537 if (dccp_hdr(rxskb
)->dccph_type
== DCCP_PKT_RESET
)
540 if (skb_rtable(rxskb
)->rt_type
!= RTN_LOCAL
)
543 dst
= dccp_v4_route_skb(net
, ctl_sk
, rxskb
);
547 skb
= dccp_ctl_make_reset(ctl_sk
, rxskb
);
551 rxiph
= ip_hdr(rxskb
);
552 dccp_hdr(skb
)->dccph_checksum
= dccp_v4_csum_finish(skb
, rxiph
->saddr
,
554 skb_dst_set(skb
, dst_clone(dst
));
556 bh_lock_sock(ctl_sk
);
557 err
= ip_build_and_send_pkt(skb
, ctl_sk
,
558 rxiph
->daddr
, rxiph
->saddr
, NULL
);
559 bh_unlock_sock(ctl_sk
);
561 if (net_xmit_eval(err
) == 0) {
562 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS
);
563 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS
);
569 static void dccp_v4_reqsk_destructor(struct request_sock
*req
)
571 dccp_feat_list_purge(&dccp_rsk(req
)->dreq_featneg
);
572 kfree(inet_rsk(req
)->opt
);
575 void dccp_syn_ack_timeout(struct sock
*sk
, struct request_sock
*req
)
578 EXPORT_SYMBOL(dccp_syn_ack_timeout
);
580 static struct request_sock_ops dccp_request_sock_ops __read_mostly
= {
582 .obj_size
= sizeof(struct dccp_request_sock
),
583 .rtx_syn_ack
= dccp_v4_send_response
,
584 .send_ack
= dccp_reqsk_send_ack
,
585 .destructor
= dccp_v4_reqsk_destructor
,
586 .send_reset
= dccp_v4_ctl_send_reset
,
587 .syn_ack_timeout
= dccp_syn_ack_timeout
,
590 int dccp_v4_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
592 struct inet_request_sock
*ireq
;
593 struct request_sock
*req
;
594 struct dccp_request_sock
*dreq
;
595 const __be32 service
= dccp_hdr_request(skb
)->dccph_req_service
;
596 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
598 /* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
599 if (skb_rtable(skb
)->rt_flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
))
600 return 0; /* discard, don't send a reset here */
602 if (dccp_bad_service_code(sk
, service
)) {
603 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_BAD_SERVICE_CODE
;
607 * TW buckets are converted to open requests without
608 * limitations, they conserve resources and peer is
609 * evidently real one.
611 dcb
->dccpd_reset_code
= DCCP_RESET_CODE_TOO_BUSY
;
612 if (inet_csk_reqsk_queue_is_full(sk
))
616 * Accept backlog is full. If we have already queued enough
617 * of warm entries in syn queue, drop request. It is better than
618 * clogging syn queue with openreqs with exponentially increasing
621 if (sk_acceptq_is_full(sk
) && inet_csk_reqsk_queue_young(sk
) > 1)
624 req
= inet_reqsk_alloc(&dccp_request_sock_ops
, sk
);
628 if (dccp_reqsk_init(req
, dccp_sk(sk
), skb
))
631 dreq
= dccp_rsk(req
);
632 if (dccp_parse_options(sk
, dreq
, skb
))
635 if (security_inet_conn_request(sk
, skb
, req
))
638 ireq
= inet_rsk(req
);
639 sk_rcv_saddr_set(req_to_sk(req
), ip_hdr(skb
)->daddr
);
640 sk_daddr_set(req_to_sk(req
), ip_hdr(skb
)->saddr
);
641 ireq
->ireq_family
= AF_INET
;
642 ireq
->ir_iif
= sk
->sk_bound_dev_if
;
645 * Step 3: Process LISTEN state
647 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
649 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
651 dreq
->dreq_isr
= dcb
->dccpd_seq
;
652 dreq
->dreq_gsr
= dreq
->dreq_isr
;
653 dreq
->dreq_iss
= dccp_v4_init_sequence(skb
);
654 dreq
->dreq_gss
= dreq
->dreq_iss
;
655 dreq
->dreq_service
= service
;
657 if (dccp_v4_send_response(sk
, req
))
660 inet_csk_reqsk_queue_hash_add(sk
, req
, DCCP_TIMEOUT_INIT
);
666 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS
);
669 EXPORT_SYMBOL_GPL(dccp_v4_conn_request
);
671 int dccp_v4_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
673 struct dccp_hdr
*dh
= dccp_hdr(skb
);
675 if (sk
->sk_state
== DCCP_OPEN
) { /* Fast path */
676 if (dccp_rcv_established(sk
, skb
, dh
, skb
->len
))
682 * Step 3: Process LISTEN state
683 * If P.type == Request or P contains a valid Init Cookie option,
684 * (* Must scan the packet's options to check for Init
685 * Cookies. Only Init Cookies are processed here,
686 * however; other options are processed in Step 8. This
687 * scan need only be performed if the endpoint uses Init
689 * (* Generate a new socket and switch to that socket *)
690 * Set S := new socket for this port pair
692 * Choose S.ISS (initial seqno) or set from Init Cookies
693 * Initialize S.GAR := S.ISS
694 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
695 * Continue with S.state == RESPOND
696 * (* A Response packet will be generated in Step 11 *)
698 * Generate Reset(No Connection) unless P.type == Reset
699 * Drop packet and return
701 * NOTE: the check for the packet types is done in
702 * dccp_rcv_state_process
704 if (sk
->sk_state
== DCCP_LISTEN
) {
705 struct sock
*nsk
= dccp_v4_hnd_req(sk
, skb
);
711 if (dccp_child_process(sk
, nsk
, skb
))
717 if (dccp_rcv_state_process(sk
, skb
, dh
, skb
->len
))
722 dccp_v4_ctl_send_reset(sk
, skb
);
727 EXPORT_SYMBOL_GPL(dccp_v4_do_rcv
);
730 * dccp_invalid_packet - check for malformed packets
731 * Implements RFC 4340, 8.5: Step 1: Check header basics
732 * Packets that fail these checks are ignored and do not receive Resets.
734 int dccp_invalid_packet(struct sk_buff
*skb
)
736 const struct dccp_hdr
*dh
;
739 if (skb
->pkt_type
!= PACKET_HOST
)
742 /* If the packet is shorter than 12 bytes, drop packet and return */
743 if (!pskb_may_pull(skb
, sizeof(struct dccp_hdr
))) {
744 DCCP_WARN("pskb_may_pull failed\n");
750 /* If P.type is not understood, drop packet and return */
751 if (dh
->dccph_type
>= DCCP_PKT_INVALID
) {
752 DCCP_WARN("invalid packet type\n");
757 * If P.Data Offset is too small for packet type, drop packet and return
759 if (dh
->dccph_doff
< dccp_hdr_len(skb
) / sizeof(u32
)) {
760 DCCP_WARN("P.Data Offset(%u) too small\n", dh
->dccph_doff
);
764 * If P.Data Offset is too too large for packet, drop packet and return
766 if (!pskb_may_pull(skb
, dh
->dccph_doff
* sizeof(u32
))) {
767 DCCP_WARN("P.Data Offset(%u) too large\n", dh
->dccph_doff
);
772 * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
773 * has short sequence numbers), drop packet and return
775 if ((dh
->dccph_type
< DCCP_PKT_DATA
||
776 dh
->dccph_type
> DCCP_PKT_DATAACK
) && dh
->dccph_x
== 0) {
777 DCCP_WARN("P.type (%s) not Data || [Data]Ack, while P.X == 0\n",
778 dccp_packet_name(dh
->dccph_type
));
783 * If P.CsCov is too large for the packet size, drop packet and return.
784 * This must come _before_ checksumming (not as RFC 4340 suggests).
786 cscov
= dccp_csum_coverage(skb
);
787 if (cscov
> skb
->len
) {
788 DCCP_WARN("P.CsCov %u exceeds packet length %d\n",
789 dh
->dccph_cscov
, skb
->len
);
793 /* If header checksum is incorrect, drop packet and return.
794 * (This step is completed in the AF-dependent functions.) */
795 skb
->csum
= skb_checksum(skb
, 0, cscov
, 0);
799 EXPORT_SYMBOL_GPL(dccp_invalid_packet
);
801 /* this is called when real data arrives */
802 static int dccp_v4_rcv(struct sk_buff
*skb
)
804 const struct dccp_hdr
*dh
;
805 const struct iphdr
*iph
;
809 /* Step 1: Check header basics */
811 if (dccp_invalid_packet(skb
))
815 /* Step 1: If header checksum is incorrect, drop packet and return */
816 if (dccp_v4_csum_finish(skb
, iph
->saddr
, iph
->daddr
)) {
817 DCCP_WARN("dropped packet with invalid checksum\n");
823 DCCP_SKB_CB(skb
)->dccpd_seq
= dccp_hdr_seq(dh
);
824 DCCP_SKB_CB(skb
)->dccpd_type
= dh
->dccph_type
;
826 dccp_pr_debug("%8.8s src=%pI4@%-5d dst=%pI4@%-5d seq=%llu",
827 dccp_packet_name(dh
->dccph_type
),
828 &iph
->saddr
, ntohs(dh
->dccph_sport
),
829 &iph
->daddr
, ntohs(dh
->dccph_dport
),
830 (unsigned long long) DCCP_SKB_CB(skb
)->dccpd_seq
);
832 if (dccp_packet_without_ack(skb
)) {
833 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= DCCP_PKT_WITHOUT_ACK_SEQ
;
834 dccp_pr_debug_cat("\n");
836 DCCP_SKB_CB(skb
)->dccpd_ack_seq
= dccp_hdr_ack_seq(skb
);
837 dccp_pr_debug_cat(", ack=%llu\n", (unsigned long long)
838 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
842 * Look up flow ID in table and get corresponding socket */
843 sk
= __inet_lookup_skb(&dccp_hashinfo
, skb
,
844 dh
->dccph_sport
, dh
->dccph_dport
);
850 dccp_pr_debug("failed to look up flow ID in table and "
851 "get corresponding socket\n");
857 * ... or S.state == TIMEWAIT,
858 * Generate Reset(No Connection) unless P.type == Reset
859 * Drop packet and return
861 if (sk
->sk_state
== DCCP_TIME_WAIT
) {
862 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
863 inet_twsk_put(inet_twsk(sk
));
868 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
869 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
870 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
872 min_cov
= dccp_sk(sk
)->dccps_pcrlen
;
873 if (dh
->dccph_cscov
&& (min_cov
== 0 || dh
->dccph_cscov
< min_cov
)) {
874 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
875 dh
->dccph_cscov
, min_cov
);
876 /* FIXME: "Such packets SHOULD be reported using Data Dropped
877 * options (Section 11.7) with Drop Code 0, Protocol
879 goto discard_and_relse
;
882 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
883 goto discard_and_relse
;
886 return sk_receive_skb(sk
, skb
, 1);
889 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
894 * Generate Reset(No Connection) unless P.type == Reset
895 * Drop packet and return
897 if (dh
->dccph_type
!= DCCP_PKT_RESET
) {
898 DCCP_SKB_CB(skb
)->dccpd_reset_code
=
899 DCCP_RESET_CODE_NO_CONNECTION
;
900 dccp_v4_ctl_send_reset(sk
, skb
);
912 static const struct inet_connection_sock_af_ops dccp_ipv4_af_ops
= {
913 .queue_xmit
= ip_queue_xmit
,
914 .send_check
= dccp_v4_send_check
,
915 .rebuild_header
= inet_sk_rebuild_header
,
916 .conn_request
= dccp_v4_conn_request
,
917 .syn_recv_sock
= dccp_v4_request_recv_sock
,
918 .net_header_len
= sizeof(struct iphdr
),
919 .setsockopt
= ip_setsockopt
,
920 .getsockopt
= ip_getsockopt
,
921 .addr2sockaddr
= inet_csk_addr2sockaddr
,
922 .sockaddr_len
= sizeof(struct sockaddr_in
),
923 .bind_conflict
= inet_csk_bind_conflict
,
925 .compat_setsockopt
= compat_ip_setsockopt
,
926 .compat_getsockopt
= compat_ip_getsockopt
,
930 static int dccp_v4_init_sock(struct sock
*sk
)
932 static __u8 dccp_v4_ctl_sock_initialized
;
933 int err
= dccp_init_sock(sk
, dccp_v4_ctl_sock_initialized
);
936 if (unlikely(!dccp_v4_ctl_sock_initialized
))
937 dccp_v4_ctl_sock_initialized
= 1;
938 inet_csk(sk
)->icsk_af_ops
= &dccp_ipv4_af_ops
;
944 static struct timewait_sock_ops dccp_timewait_sock_ops
= {
945 .twsk_obj_size
= sizeof(struct inet_timewait_sock
),
948 static struct proto dccp_v4_prot
= {
950 .owner
= THIS_MODULE
,
952 .connect
= dccp_v4_connect
,
953 .disconnect
= dccp_disconnect
,
955 .init
= dccp_v4_init_sock
,
956 .setsockopt
= dccp_setsockopt
,
957 .getsockopt
= dccp_getsockopt
,
958 .sendmsg
= dccp_sendmsg
,
959 .recvmsg
= dccp_recvmsg
,
960 .backlog_rcv
= dccp_v4_do_rcv
,
962 .unhash
= inet_unhash
,
963 .accept
= inet_csk_accept
,
964 .get_port
= inet_csk_get_port
,
965 .shutdown
= dccp_shutdown
,
966 .destroy
= dccp_destroy_sock
,
967 .orphan_count
= &dccp_orphan_count
,
968 .max_header
= MAX_DCCP_HEADER
,
969 .obj_size
= sizeof(struct dccp_sock
),
970 .slab_flags
= SLAB_DESTROY_BY_RCU
,
971 .rsk_prot
= &dccp_request_sock_ops
,
972 .twsk_prot
= &dccp_timewait_sock_ops
,
973 .h
.hashinfo
= &dccp_hashinfo
,
975 .compat_setsockopt
= compat_dccp_setsockopt
,
976 .compat_getsockopt
= compat_dccp_getsockopt
,
980 static const struct net_protocol dccp_v4_protocol
= {
981 .handler
= dccp_v4_rcv
,
982 .err_handler
= dccp_v4_err
,
985 .icmp_strict_tag_validation
= 1,
988 static const struct proto_ops inet_dccp_ops
= {
990 .owner
= THIS_MODULE
,
991 .release
= inet_release
,
993 .connect
= inet_stream_connect
,
994 .socketpair
= sock_no_socketpair
,
995 .accept
= inet_accept
,
996 .getname
= inet_getname
,
997 /* FIXME: work on tcp_poll to rename it to inet_csk_poll */
1000 /* FIXME: work on inet_listen to rename it to sock_common_listen */
1001 .listen
= inet_dccp_listen
,
1002 .shutdown
= inet_shutdown
,
1003 .setsockopt
= sock_common_setsockopt
,
1004 .getsockopt
= sock_common_getsockopt
,
1005 .sendmsg
= inet_sendmsg
,
1006 .recvmsg
= sock_common_recvmsg
,
1007 .mmap
= sock_no_mmap
,
1008 .sendpage
= sock_no_sendpage
,
1009 #ifdef CONFIG_COMPAT
1010 .compat_setsockopt
= compat_sock_common_setsockopt
,
1011 .compat_getsockopt
= compat_sock_common_getsockopt
,
1015 static struct inet_protosw dccp_v4_protosw
= {
1017 .protocol
= IPPROTO_DCCP
,
1018 .prot
= &dccp_v4_prot
,
1019 .ops
= &inet_dccp_ops
,
1020 .flags
= INET_PROTOSW_ICSK
,
1023 static int __net_init
dccp_v4_init_net(struct net
*net
)
1025 if (dccp_hashinfo
.bhash
== NULL
)
1026 return -ESOCKTNOSUPPORT
;
1028 return inet_ctl_sock_create(&net
->dccp
.v4_ctl_sk
, PF_INET
,
1029 SOCK_DCCP
, IPPROTO_DCCP
, net
);
1032 static void __net_exit
dccp_v4_exit_net(struct net
*net
)
1034 inet_ctl_sock_destroy(net
->dccp
.v4_ctl_sk
);
1037 static struct pernet_operations dccp_v4_ops
= {
1038 .init
= dccp_v4_init_net
,
1039 .exit
= dccp_v4_exit_net
,
1042 static int __init
dccp_v4_init(void)
1044 int err
= proto_register(&dccp_v4_prot
, 1);
1049 err
= inet_add_protocol(&dccp_v4_protocol
, IPPROTO_DCCP
);
1051 goto out_proto_unregister
;
1053 inet_register_protosw(&dccp_v4_protosw
);
1055 err
= register_pernet_subsys(&dccp_v4_ops
);
1057 goto out_destroy_ctl_sock
;
1060 out_destroy_ctl_sock
:
1061 inet_unregister_protosw(&dccp_v4_protosw
);
1062 inet_del_protocol(&dccp_v4_protocol
, IPPROTO_DCCP
);
1063 out_proto_unregister
:
1064 proto_unregister(&dccp_v4_prot
);
1068 static void __exit
dccp_v4_exit(void)
1070 unregister_pernet_subsys(&dccp_v4_ops
);
1071 inet_unregister_protosw(&dccp_v4_protosw
);
1072 inet_del_protocol(&dccp_v4_protocol
, IPPROTO_DCCP
);
1073 proto_unregister(&dccp_v4_prot
);
1076 module_init(dccp_v4_init
);
1077 module_exit(dccp_v4_exit
);
1080 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1081 * values directly, Also cover the case where the protocol is not specified,
1082 * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
1084 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 33, 6);
1085 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET
, 0, 6);
1086 MODULE_LICENSE("GPL");
1087 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1088 MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");