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/config.h>
14 #include <linux/dccp.h>
15 #include <linux/icmp.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/random.h>
21 #include <net/inet_hashtables.h>
22 #include <net/inet_sock.h>
24 #include <net/timewait_sock.h>
25 #include <net/tcp_states.h>
33 struct inet_hashinfo __cacheline_aligned dccp_hashinfo = {
34 .lhash_lock = RW_LOCK_UNLOCKED,
35 .lhash_users = ATOMIC_INIT(0),
36 .lhash_wait = __WAIT_QUEUE_HEAD_INITIALIZER(dccp_hashinfo.lhash_wait),
39 EXPORT_SYMBOL_GPL(dccp_hashinfo);
41 static int dccp_v4_get_port(struct sock *sk, const unsigned short snum)
43 return inet_csk_get_port(&dccp_hashinfo, sk, snum,
44 inet_csk_bind_conflict);
47 static void dccp_v4_hash(struct sock *sk)
49 inet_hash(&dccp_hashinfo, sk);
52 void dccp_unhash(struct sock *sk)
54 inet_unhash(&dccp_hashinfo, sk);
57 EXPORT_SYMBOL_GPL(dccp_unhash);
59 int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
61 struct inet_sock *inet = inet_sk(sk);
62 struct dccp_sock *dp = dccp_sk(sk);
63 const struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
69 dp->dccps_role = DCCP_ROLE_CLIENT;
71 if (dccp_service_not_initialized(sk))
74 if (addr_len < sizeof(struct sockaddr_in))
77 if (usin->sin_family != AF_INET)
80 nexthop = daddr = usin->sin_addr.s_addr;
81 if (inet->opt != NULL && inet->opt->srr) {
84 nexthop = inet->opt->faddr;
87 tmp = ip_route_connect(&rt, nexthop, inet->saddr,
88 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
90 inet->sport, usin->sin_port, sk);
94 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
99 if (inet->opt == NULL || !inet->opt->srr)
102 if (inet->saddr == 0)
103 inet->saddr = rt->rt_src;
104 inet->rcv_saddr = inet->saddr;
106 inet->dport = usin->sin_port;
109 inet_csk(sk)->icsk_ext_hdr_len = 0;
110 if (inet->opt != NULL)
111 inet_csk(sk)->icsk_ext_hdr_len = inet->opt->optlen;
113 * Socket identity is still unknown (sport may be zero).
114 * However we set state to DCCP_REQUESTING and not releasing socket
115 * lock select source port, enter ourselves into the hash tables and
116 * complete initialization after this.
118 dccp_set_state(sk, DCCP_REQUESTING);
119 err = inet_hash_connect(&dccp_death_row, sk);
123 err = ip_route_newports(&rt, IPPROTO_DCCP, inet->sport, inet->dport,
128 /* OK, now commit destination to socket. */
129 sk_setup_caps(sk, &rt->u.dst);
132 dp->dccps_iss = secure_dccp_sequence_number(inet->saddr,
136 dccp_update_gss(sk, dp->dccps_iss);
138 inet->id = dp->dccps_iss ^ jiffies;
140 err = dccp_connect(sk);
148 * This unhashes the socket and releases the local port, if necessary.
150 dccp_set_state(sk, DCCP_CLOSED);
152 sk->sk_route_caps = 0;
157 EXPORT_SYMBOL_GPL(dccp_v4_connect);
160 * This routine does path mtu discovery as defined in RFC1191.
162 static inline void dccp_do_pmtu_discovery(struct sock *sk,
163 const struct iphdr *iph,
166 struct dst_entry *dst;
167 const struct inet_sock *inet = inet_sk(sk);
168 const struct dccp_sock *dp = dccp_sk(sk);
170 /* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
171 * send out by Linux are always < 576bytes so they should go through
174 if (sk->sk_state == DCCP_LISTEN)
177 /* We don't check in the destentry if pmtu discovery is forbidden
178 * on this route. We just assume that no packet_to_big packets
179 * are send back when pmtu discovery is not active.
180 * There is a small race when the user changes this flag in the
181 * route, but I think that's acceptable.
183 if ((dst = __sk_dst_check(sk, 0)) == NULL)
186 dst->ops->update_pmtu(dst, mtu);
188 /* Something is about to be wrong... Remember soft error
189 * for the case, if this connection will not able to recover.
191 if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
192 sk->sk_err_soft = EMSGSIZE;
196 if (inet->pmtudisc != IP_PMTUDISC_DONT &&
197 inet_csk(sk)->icsk_pmtu_cookie > mtu) {
198 dccp_sync_mss(sk, mtu);
201 * From: draft-ietf-dccp-spec-11.txt
203 * DCCP-Sync packets are the best choice for upward
204 * probing, since DCCP-Sync probes do not risk application
207 dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC);
208 } /* else let the usual retransmit timer handle it */
211 static void dccp_v4_ctl_send_ack(struct sk_buff *rxskb)
214 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
215 const int dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
216 sizeof(struct dccp_hdr_ext) +
217 sizeof(struct dccp_hdr_ack_bits);
220 if (((struct rtable *)rxskb->dst)->rt_type != RTN_LOCAL)
223 skb = alloc_skb(MAX_DCCP_HEADER + 15, GFP_ATOMIC);
227 /* Reserve space for headers. */
228 skb_reserve(skb, MAX_DCCP_HEADER);
230 skb->dst = dst_clone(rxskb->dst);
232 skb->h.raw = skb_push(skb, dccp_hdr_ack_len);
234 memset(dh, 0, dccp_hdr_ack_len);
236 /* Build DCCP header and checksum it. */
237 dh->dccph_type = DCCP_PKT_ACK;
238 dh->dccph_sport = rxdh->dccph_dport;
239 dh->dccph_dport = rxdh->dccph_sport;
240 dh->dccph_doff = dccp_hdr_ack_len / 4;
243 dccp_hdr_set_seq(dh, DCCP_SKB_CB(rxskb)->dccpd_ack_seq);
244 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
245 DCCP_SKB_CB(rxskb)->dccpd_seq);
247 bh_lock_sock(dccp_ctl_socket->sk);
248 err = ip_build_and_send_pkt(skb, dccp_ctl_socket->sk,
249 rxskb->nh.iph->daddr,
250 rxskb->nh.iph->saddr, NULL);
251 bh_unlock_sock(dccp_ctl_socket->sk);
253 if (err == NET_XMIT_CN || err == 0) {
254 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
255 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
259 static void dccp_v4_reqsk_send_ack(struct sk_buff *skb,
260 struct request_sock *req)
262 dccp_v4_ctl_send_ack(skb);
265 static int dccp_v4_send_response(struct sock *sk, struct request_sock *req,
266 struct dst_entry *dst)
271 /* First, grab a route. */
273 if (dst == NULL && (dst = inet_csk_route_req(sk, req)) == NULL)
276 skb = dccp_make_response(sk, dst, req);
278 const struct inet_request_sock *ireq = inet_rsk(req);
280 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
281 err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr,
284 if (err == NET_XMIT_CN)
294 * This routine is called by the ICMP module when it gets some sort of error
295 * condition. If err < 0 then the socket should be closed and the error
296 * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
297 * After adjustment header points to the first 8 bytes of the tcp header. We
298 * need to find the appropriate port.
300 * The locking strategy used here is very "optimistic". When someone else
301 * accesses the socket the ICMP is just dropped and for some paths there is no
302 * check at all. A more general error queue to queue errors for later handling
303 * is probably better.
305 void dccp_v4_err(struct sk_buff *skb, u32 info)
307 const struct iphdr *iph = (struct iphdr *)skb->data;
308 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data +
310 struct dccp_sock *dp;
311 struct inet_sock *inet;
312 const int type = skb->h.icmph->type;
313 const int code = skb->h.icmph->code;
318 if (skb->len < (iph->ihl << 2) + 8) {
319 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
323 sk = inet_lookup(&dccp_hashinfo, iph->daddr, dh->dccph_dport,
324 iph->saddr, dh->dccph_sport, inet_iif(skb));
326 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
330 if (sk->sk_state == DCCP_TIME_WAIT) {
331 inet_twsk_put((struct inet_timewait_sock *)sk);
336 /* If too many ICMPs get dropped on busy
337 * servers this needs to be solved differently.
339 if (sock_owned_by_user(sk))
340 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
342 if (sk->sk_state == DCCP_CLOSED)
346 seq = dccp_hdr_seq(skb);
347 if (sk->sk_state != DCCP_LISTEN &&
348 !between48(seq, dp->dccps_swl, dp->dccps_swh)) {
349 NET_INC_STATS(LINUX_MIB_OUTOFWINDOWICMPS);
354 case ICMP_SOURCE_QUENCH:
355 /* Just silently ignore these. */
357 case ICMP_PARAMETERPROB:
360 case ICMP_DEST_UNREACH:
361 if (code > NR_ICMP_UNREACH)
364 if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
365 if (!sock_owned_by_user(sk))
366 dccp_do_pmtu_discovery(sk, iph, info);
370 err = icmp_err_convert[code].errno;
372 case ICMP_TIME_EXCEEDED:
379 switch (sk->sk_state) {
380 struct request_sock *req , **prev;
382 if (sock_owned_by_user(sk))
384 req = inet_csk_search_req(sk, &prev, dh->dccph_dport,
385 iph->daddr, iph->saddr);
390 * ICMPs are not backlogged, hence we cannot get an established
395 if (seq != dccp_rsk(req)->dreq_iss) {
396 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
400 * Still in RESPOND, just remove it silently.
401 * There is no good way to pass the error to the newly
402 * created socket, and POSIX does not want network
403 * errors returned from accept().
405 inet_csk_reqsk_queue_drop(sk, req, prev);
408 case DCCP_REQUESTING:
410 if (!sock_owned_by_user(sk)) {
411 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
414 sk->sk_error_report(sk);
418 sk->sk_err_soft = err;
422 /* If we've already connected we will keep trying
423 * until we time out, or the user gives up.
425 * rfc1122 4.2.3.9 allows to consider as hard errors
426 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
427 * but it is obsoleted by pmtu discovery).
429 * Note, that in modern internet, where routing is unreliable
430 * and in each dark corner broken firewalls sit, sending random
431 * errors ordered by their masters even this two messages finally lose
432 * their original sense (even Linux sends invalid PORT_UNREACHs)
434 * Now we are in compliance with RFCs.
439 if (!sock_owned_by_user(sk) && inet->recverr) {
441 sk->sk_error_report(sk);
442 } else /* Only an error on timeout */
443 sk->sk_err_soft = err;
449 /* This routine computes an IPv4 DCCP checksum. */
450 void dccp_v4_send_check(struct sock *sk, int len, struct sk_buff *skb)
452 const struct inet_sock *inet = inet_sk(sk);
453 struct dccp_hdr *dh = dccp_hdr(skb);
455 dh->dccph_checksum = dccp_v4_checksum(skb, inet->saddr, inet->daddr);
458 EXPORT_SYMBOL_GPL(dccp_v4_send_check);
460 int dccp_v4_send_reset(struct sock *sk, enum dccp_reset_codes code)
464 * FIXME: what if rebuild_header fails?
465 * Should we be doing a rebuild_header here?
467 int err = inet_sk_rebuild_header(sk);
472 skb = dccp_make_reset(sk, sk->sk_dst_cache, code);
474 const struct inet_sock *inet = inet_sk(sk);
476 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
477 err = ip_build_and_send_pkt(skb, sk,
478 inet->saddr, inet->daddr, NULL);
479 if (err == NET_XMIT_CN)
486 static inline u64 dccp_v4_init_sequence(const struct sock *sk,
487 const struct sk_buff *skb)
489 return secure_dccp_sequence_number(skb->nh.iph->daddr,
491 dccp_hdr(skb)->dccph_dport,
492 dccp_hdr(skb)->dccph_sport);
495 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
497 struct inet_request_sock *ireq;
499 struct request_sock *req;
500 struct dccp_request_sock *dreq;
501 const __u32 saddr = skb->nh.iph->saddr;
502 const __u32 daddr = skb->nh.iph->daddr;
503 const __u32 service = dccp_hdr_request(skb)->dccph_req_service;
504 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
505 __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
507 /* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
508 if (((struct rtable *)skb->dst)->rt_flags &
509 (RTCF_BROADCAST | RTCF_MULTICAST)) {
510 reset_code = DCCP_RESET_CODE_NO_CONNECTION;
514 if (dccp_bad_service_code(sk, service)) {
515 reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
519 * TW buckets are converted to open requests without
520 * limitations, they conserve resources and peer is
521 * evidently real one.
523 if (inet_csk_reqsk_queue_is_full(sk))
527 * Accept backlog is full. If we have already queued enough
528 * of warm entries in syn queue, drop request. It is better than
529 * clogging syn queue with openreqs with exponentially increasing
532 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
535 req = reqsk_alloc(sk->sk_prot->rsk_prot);
539 if (dccp_parse_options(sk, skb))
542 dccp_openreq_init(req, &dp, skb);
544 ireq = inet_rsk(req);
545 ireq->loc_addr = daddr;
546 ireq->rmt_addr = saddr;
547 req->rcv_wnd = 100; /* Fake, option parsing will get the
552 * Step 3: Process LISTEN state
554 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
556 * In fact we defer setting S.GSR, S.SWL, S.SWH to
557 * dccp_create_openreq_child.
559 dreq = dccp_rsk(req);
560 dreq->dreq_isr = dcb->dccpd_seq;
561 dreq->dreq_iss = dccp_v4_init_sequence(sk, skb);
562 dreq->dreq_service = service;
564 if (dccp_v4_send_response(sk, req, NULL))
567 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
573 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
574 dcb->dccpd_reset_code = reset_code;
578 EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
581 * The three way handshake has completed - we got a valid ACK or DATAACK -
582 * now create the new socket.
584 * This is the equivalent of TCP's tcp_v4_syn_recv_sock
586 struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
587 struct request_sock *req,
588 struct dst_entry *dst)
590 struct inet_request_sock *ireq;
591 struct inet_sock *newinet;
592 struct dccp_sock *newdp;
595 if (sk_acceptq_is_full(sk))
598 if (dst == NULL && (dst = inet_csk_route_req(sk, req)) == NULL)
601 newsk = dccp_create_openreq_child(sk, req, skb);
605 sk_setup_caps(newsk, dst);
607 newdp = dccp_sk(newsk);
608 newinet = inet_sk(newsk);
609 ireq = inet_rsk(req);
610 newinet->daddr = ireq->rmt_addr;
611 newinet->rcv_saddr = ireq->loc_addr;
612 newinet->saddr = ireq->loc_addr;
613 newinet->opt = ireq->opt;
615 newinet->mc_index = inet_iif(skb);
616 newinet->mc_ttl = skb->nh.iph->ttl;
617 newinet->id = jiffies;
619 dccp_sync_mss(newsk, dst_mtu(dst));
621 __inet_hash(&dccp_hashinfo, newsk, 0);
622 __inet_inherit_port(&dccp_hashinfo, sk, newsk);
627 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
629 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
634 EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
636 static struct sock *dccp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
638 const struct dccp_hdr *dh = dccp_hdr(skb);
639 const struct iphdr *iph = skb->nh.iph;
641 struct request_sock **prev;
642 /* Find possible connection requests. */
643 struct request_sock *req = inet_csk_search_req(sk, &prev,
645 iph->saddr, iph->daddr);
647 return dccp_check_req(sk, skb, req, prev);
649 nsk = __inet_lookup_established(&dccp_hashinfo,
650 iph->saddr, dh->dccph_sport,
651 iph->daddr, ntohs(dh->dccph_dport),
654 if (nsk->sk_state != DCCP_TIME_WAIT) {
658 inet_twsk_put((struct inet_timewait_sock *)nsk);
665 int dccp_v4_checksum(const struct sk_buff *skb, const u32 saddr,
668 const struct dccp_hdr* dh = dccp_hdr(skb);
672 if (dh->dccph_cscov == 0)
673 checksum_len = skb->len;
675 checksum_len = (dh->dccph_cscov + dh->dccph_x) * sizeof(u32);
676 checksum_len = checksum_len < skb->len ? checksum_len :
680 tmp = csum_partial((unsigned char *)dh, checksum_len, 0);
681 return csum_tcpudp_magic(saddr, daddr, checksum_len,
685 static int dccp_v4_verify_checksum(struct sk_buff *skb,
686 const u32 saddr, const u32 daddr)
688 struct dccp_hdr *dh = dccp_hdr(skb);
692 if (dh->dccph_cscov == 0)
693 checksum_len = skb->len;
695 checksum_len = (dh->dccph_cscov + dh->dccph_x) * sizeof(u32);
696 checksum_len = checksum_len < skb->len ? checksum_len :
699 tmp = csum_partial((unsigned char *)dh, checksum_len, 0);
700 return csum_tcpudp_magic(saddr, daddr, checksum_len,
701 IPPROTO_DCCP, tmp) == 0 ? 0 : -1;
704 static struct dst_entry* dccp_v4_route_skb(struct sock *sk,
708 struct flowi fl = { .oif = ((struct rtable *)skb->dst)->rt_iif,
710 { .daddr = skb->nh.iph->saddr,
711 .saddr = skb->nh.iph->daddr,
712 .tos = RT_CONN_FLAGS(sk) } },
713 .proto = sk->sk_protocol,
715 { .sport = dccp_hdr(skb)->dccph_dport,
716 .dport = dccp_hdr(skb)->dccph_sport }
720 if (ip_route_output_flow(&rt, &fl, sk, 0)) {
721 IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES);
728 static void dccp_v4_ctl_send_reset(struct sk_buff *rxskb)
731 struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
732 const int dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
733 sizeof(struct dccp_hdr_ext) +
734 sizeof(struct dccp_hdr_reset);
736 struct dst_entry *dst;
739 /* Never send a reset in response to a reset. */
740 if (rxdh->dccph_type == DCCP_PKT_RESET)
743 if (((struct rtable *)rxskb->dst)->rt_type != RTN_LOCAL)
746 dst = dccp_v4_route_skb(dccp_ctl_socket->sk, rxskb);
750 skb = alloc_skb(MAX_DCCP_HEADER + 15, GFP_ATOMIC);
754 /* Reserve space for headers. */
755 skb_reserve(skb, MAX_DCCP_HEADER);
756 skb->dst = dst_clone(dst);
758 skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
760 memset(dh, 0, dccp_hdr_reset_len);
762 /* Build DCCP header and checksum it. */
763 dh->dccph_type = DCCP_PKT_RESET;
764 dh->dccph_sport = rxdh->dccph_dport;
765 dh->dccph_dport = rxdh->dccph_sport;
766 dh->dccph_doff = dccp_hdr_reset_len / 4;
768 dccp_hdr_reset(skb)->dccph_reset_code =
769 DCCP_SKB_CB(rxskb)->dccpd_reset_code;
771 /* See "8.3.1. Abnormal Termination" in draft-ietf-dccp-spec-11 */
773 if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
774 dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
776 dccp_hdr_set_seq(dh, seqno);
777 dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
778 DCCP_SKB_CB(rxskb)->dccpd_seq);
780 dh->dccph_checksum = dccp_v4_checksum(skb, rxskb->nh.iph->saddr,
781 rxskb->nh.iph->daddr);
783 bh_lock_sock(dccp_ctl_socket->sk);
784 err = ip_build_and_send_pkt(skb, dccp_ctl_socket->sk,
785 rxskb->nh.iph->daddr,
786 rxskb->nh.iph->saddr, NULL);
787 bh_unlock_sock(dccp_ctl_socket->sk);
789 if (err == NET_XMIT_CN || err == 0) {
790 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
791 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
797 int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
799 struct dccp_hdr *dh = dccp_hdr(skb);
801 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
802 if (dccp_rcv_established(sk, skb, dh, skb->len))
808 * Step 3: Process LISTEN state
809 * If S.state == LISTEN,
810 * If P.type == Request or P contains a valid Init Cookie
812 * * Must scan the packet's options to check for an Init
813 * Cookie. Only the Init Cookie is processed here,
814 * however; other options are processed in Step 8. This
815 * scan need only be performed if the endpoint uses Init
817 * * Generate a new socket and switch to that socket *
818 * Set S := new socket for this port pair
820 * Choose S.ISS (initial seqno) or set from Init Cookie
821 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
822 * Continue with S.state == RESPOND
823 * * A Response packet will be generated in Step 11 *
825 * Generate Reset(No Connection) unless P.type == Reset
826 * Drop packet and return
828 * NOTE: the check for the packet types is done in
829 * dccp_rcv_state_process
831 if (sk->sk_state == DCCP_LISTEN) {
832 struct sock *nsk = dccp_v4_hnd_req(sk, skb);
838 if (dccp_child_process(sk, nsk, skb))
844 if (dccp_rcv_state_process(sk, skb, dh, skb->len))
849 dccp_v4_ctl_send_reset(skb);
855 EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
857 int dccp_invalid_packet(struct sk_buff *skb)
859 const struct dccp_hdr *dh;
861 if (skb->pkt_type != PACKET_HOST)
864 if (!pskb_may_pull(skb, sizeof(struct dccp_hdr))) {
865 LIMIT_NETDEBUG(KERN_WARNING "DCCP: pskb_may_pull failed\n");
871 /* If the packet type is not understood, drop packet and return */
872 if (dh->dccph_type >= DCCP_PKT_INVALID) {
873 LIMIT_NETDEBUG(KERN_WARNING "DCCP: invalid packet type\n");
878 * If P.Data Offset is too small for packet type, or too large for
879 * packet, drop packet and return
881 if (dh->dccph_doff < dccp_hdr_len(skb) / sizeof(u32)) {
882 LIMIT_NETDEBUG(KERN_WARNING "DCCP: P.Data Offset(%u) "
888 if (!pskb_may_pull(skb, dh->dccph_doff * sizeof(u32))) {
889 LIMIT_NETDEBUG(KERN_WARNING "DCCP: P.Data Offset(%u) "
898 * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
899 * has short sequence numbers), drop packet and return
901 if (dh->dccph_x == 0 &&
902 dh->dccph_type != DCCP_PKT_DATA &&
903 dh->dccph_type != DCCP_PKT_ACK &&
904 dh->dccph_type != DCCP_PKT_DATAACK) {
905 LIMIT_NETDEBUG(KERN_WARNING "DCCP: P.type (%s) not Data, Ack "
906 "nor DataAck and P.X == 0\n",
907 dccp_packet_name(dh->dccph_type));
914 EXPORT_SYMBOL_GPL(dccp_invalid_packet);
916 /* this is called when real data arrives */
917 int dccp_v4_rcv(struct sk_buff *skb)
919 const struct dccp_hdr *dh;
922 /* Step 1: Check header basics: */
924 if (dccp_invalid_packet(skb))
927 /* If the header checksum is incorrect, drop packet and return */
928 if (dccp_v4_verify_checksum(skb, skb->nh.iph->saddr,
929 skb->nh.iph->daddr) < 0) {
930 LIMIT_NETDEBUG(KERN_WARNING "%s: incorrect header checksum\n",
937 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb);
938 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
940 dccp_pr_debug("%8.8s "
941 "src=%u.%u.%u.%u@%-5d "
942 "dst=%u.%u.%u.%u@%-5d seq=%llu",
943 dccp_packet_name(dh->dccph_type),
944 NIPQUAD(skb->nh.iph->saddr), ntohs(dh->dccph_sport),
945 NIPQUAD(skb->nh.iph->daddr), ntohs(dh->dccph_dport),
946 (unsigned long long) DCCP_SKB_CB(skb)->dccpd_seq);
948 if (dccp_packet_without_ack(skb)) {
949 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
950 dccp_pr_debug_cat("\n");
952 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
953 dccp_pr_debug_cat(", ack=%llu\n",
955 DCCP_SKB_CB(skb)->dccpd_ack_seq);
959 * Look up flow ID in table and get corresponding socket */
960 sk = __inet_lookup(&dccp_hashinfo,
961 skb->nh.iph->saddr, dh->dccph_sport,
962 skb->nh.iph->daddr, ntohs(dh->dccph_dport),
968 * Generate Reset(No Connection) unless P.type == Reset
969 * Drop packet and return
972 dccp_pr_debug("failed to look up flow ID in table and "
973 "get corresponding socket\n");
979 * ... or S.state == TIMEWAIT,
980 * Generate Reset(No Connection) unless P.type == Reset
981 * Drop packet and return
984 if (sk->sk_state == DCCP_TIME_WAIT) {
985 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: "
990 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
991 goto discard_and_relse;
994 return sk_receive_skb(sk, skb);
997 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1001 * Generate Reset(No Connection) unless P.type == Reset
1002 * Drop packet and return
1004 if (dh->dccph_type != DCCP_PKT_RESET) {
1005 DCCP_SKB_CB(skb)->dccpd_reset_code =
1006 DCCP_RESET_CODE_NO_CONNECTION;
1007 dccp_v4_ctl_send_reset(skb);
1011 /* Discard frame. */
1020 inet_twsk_put((struct inet_timewait_sock *)sk);
1021 goto no_dccp_socket;
1024 struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
1025 .queue_xmit = ip_queue_xmit,
1026 .send_check = dccp_v4_send_check,
1027 .rebuild_header = inet_sk_rebuild_header,
1028 .conn_request = dccp_v4_conn_request,
1029 .syn_recv_sock = dccp_v4_request_recv_sock,
1030 .net_header_len = sizeof(struct iphdr),
1031 .setsockopt = ip_setsockopt,
1032 .getsockopt = ip_getsockopt,
1033 .addr2sockaddr = inet_csk_addr2sockaddr,
1034 .sockaddr_len = sizeof(struct sockaddr_in),
1037 int dccp_v4_init_sock(struct sock *sk)
1039 struct dccp_sock *dp = dccp_sk(sk);
1040 struct inet_connection_sock *icsk = inet_csk(sk);
1041 static int dccp_ctl_socket_init = 1;
1043 dccp_options_init(&dp->dccps_options);
1044 do_gettimeofday(&dp->dccps_epoch);
1047 * FIXME: We're hardcoding the CCID, and doing this at this point makes
1048 * the listening (master) sock get CCID control blocks, which is not
1049 * necessary, but for now, to not mess with the test userspace apps,
1050 * lets leave it here, later the real solution is to do this in a
1051 * setsockopt(CCIDs-I-want/accept). -acme
1053 if (likely(!dccp_ctl_socket_init)) {
1056 if (dp->dccps_options.dccpo_send_ack_vector) {
1057 dp->dccps_hc_rx_ackvec = dccp_ackvec_alloc(GFP_KERNEL);
1058 if (dp->dccps_hc_rx_ackvec == NULL)
1061 dp->dccps_hc_rx_ccid =
1062 ccid_hc_rx_new(dp->dccps_options.dccpo_rx_ccid,
1064 dp->dccps_hc_tx_ccid =
1065 ccid_hc_tx_new(dp->dccps_options.dccpo_tx_ccid,
1067 if (unlikely(dp->dccps_hc_rx_ccid == NULL ||
1068 dp->dccps_hc_tx_ccid == NULL)) {
1069 ccid_hc_rx_delete(dp->dccps_hc_rx_ccid, sk);
1070 ccid_hc_tx_delete(dp->dccps_hc_tx_ccid, sk);
1071 if (dp->dccps_options.dccpo_send_ack_vector) {
1072 dccp_ackvec_free(dp->dccps_hc_rx_ackvec);
1073 dp->dccps_hc_rx_ackvec = NULL;
1075 dp->dccps_hc_rx_ccid = dp->dccps_hc_tx_ccid = NULL;
1079 rc = dccp_feat_init(sk);
1083 /* control socket doesn't need feat nego */
1084 INIT_LIST_HEAD(&dp->dccps_options.dccpo_pending);
1085 INIT_LIST_HEAD(&dp->dccps_options.dccpo_conf);
1086 dccp_ctl_socket_init = 0;
1089 dccp_init_xmit_timers(sk);
1090 icsk->icsk_rto = DCCP_TIMEOUT_INIT;
1091 sk->sk_state = DCCP_CLOSED;
1092 sk->sk_write_space = dccp_write_space;
1093 icsk->icsk_af_ops = &dccp_ipv4_af_ops;
1094 icsk->icsk_sync_mss = dccp_sync_mss;
1095 dp->dccps_mss_cache = 536;
1096 dp->dccps_role = DCCP_ROLE_UNDEFINED;
1097 dp->dccps_service = DCCP_SERVICE_INVALID_VALUE;
1098 dp->dccps_l_ack_ratio = dp->dccps_r_ack_ratio = 1;
1103 EXPORT_SYMBOL_GPL(dccp_v4_init_sock);
1105 int dccp_v4_destroy_sock(struct sock *sk)
1107 struct dccp_sock *dp = dccp_sk(sk);
1110 * DCCP doesn't use sk_write_queue, just sk_send_head
1111 * for retransmissions
1113 if (sk->sk_send_head != NULL) {
1114 kfree_skb(sk->sk_send_head);
1115 sk->sk_send_head = NULL;
1118 /* Clean up a referenced DCCP bind bucket. */
1119 if (inet_csk(sk)->icsk_bind_hash != NULL)
1120 inet_put_port(&dccp_hashinfo, sk);
1122 kfree(dp->dccps_service_list);
1123 dp->dccps_service_list = NULL;
1125 if (dp->dccps_options.dccpo_send_ack_vector) {
1126 dccp_ackvec_free(dp->dccps_hc_rx_ackvec);
1127 dp->dccps_hc_rx_ackvec = NULL;
1129 ccid_hc_rx_delete(dp->dccps_hc_rx_ccid, sk);
1130 ccid_hc_tx_delete(dp->dccps_hc_tx_ccid, sk);
1131 dp->dccps_hc_rx_ccid = dp->dccps_hc_tx_ccid = NULL;
1133 /* clean up feature negotiation state */
1134 dccp_feat_clean(sk);
1139 EXPORT_SYMBOL_GPL(dccp_v4_destroy_sock);
1141 static void dccp_v4_reqsk_destructor(struct request_sock *req)
1143 kfree(inet_rsk(req)->opt);
1146 static struct request_sock_ops dccp_request_sock_ops = {
1148 .obj_size = sizeof(struct dccp_request_sock),
1149 .rtx_syn_ack = dccp_v4_send_response,
1150 .send_ack = dccp_v4_reqsk_send_ack,
1151 .destructor = dccp_v4_reqsk_destructor,
1152 .send_reset = dccp_v4_ctl_send_reset,
1155 static struct timewait_sock_ops dccp_timewait_sock_ops = {
1156 .twsk_obj_size = sizeof(struct inet_timewait_sock),
1159 struct proto dccp_prot = {
1161 .owner = THIS_MODULE,
1162 .close = dccp_close,
1163 .connect = dccp_v4_connect,
1164 .disconnect = dccp_disconnect,
1165 .ioctl = dccp_ioctl,
1166 .init = dccp_v4_init_sock,
1167 .setsockopt = dccp_setsockopt,
1168 .getsockopt = dccp_getsockopt,
1169 .sendmsg = dccp_sendmsg,
1170 .recvmsg = dccp_recvmsg,
1171 .backlog_rcv = dccp_v4_do_rcv,
1172 .hash = dccp_v4_hash,
1173 .unhash = dccp_unhash,
1174 .accept = inet_csk_accept,
1175 .get_port = dccp_v4_get_port,
1176 .shutdown = dccp_shutdown,
1177 .destroy = dccp_v4_destroy_sock,
1178 .orphan_count = &dccp_orphan_count,
1179 .max_header = MAX_DCCP_HEADER,
1180 .obj_size = sizeof(struct dccp_sock),
1181 .rsk_prot = &dccp_request_sock_ops,
1182 .twsk_prot = &dccp_timewait_sock_ops,
1185 EXPORT_SYMBOL_GPL(dccp_prot);