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/skbuff.h>
23 static void dccp_fin(struct sock *sk, struct sk_buff *skb)
25 sk->sk_shutdown |= RCV_SHUTDOWN;
26 sock_set_flag(sk, SOCK_DONE);
27 __skb_pull(skb, dccp_hdr(skb)->dccph_doff * 4);
28 __skb_queue_tail(&sk->sk_receive_queue, skb);
29 skb_set_owner_r(skb, sk);
30 sk->sk_data_ready(sk, 0);
33 static void dccp_rcv_close(struct sock *sk, struct sk_buff *skb)
35 dccp_v4_send_reset(sk, DCCP_RESET_CODE_CLOSED);
37 dccp_set_state(sk, DCCP_CLOSED);
38 sk_wake_async(sk, 1, POLL_HUP);
41 static void dccp_rcv_closereq(struct sock *sk, struct sk_buff *skb)
44 * Step 7: Check for unexpected packet types
45 * If (S.is_server and P.type == CloseReq)
46 * Send Sync packet acknowledging P.seqno
47 * Drop packet and return
49 if (dccp_sk(sk)->dccps_role != DCCP_ROLE_CLIENT) {
50 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq, DCCP_PKT_SYNC);
54 if (sk->sk_state != DCCP_CLOSING)
55 dccp_set_state(sk, DCCP_CLOSING);
56 dccp_send_close(sk, 0);
59 static inline void dccp_event_ack_recv(struct sock *sk, struct sk_buff *skb)
61 struct dccp_sock *dp = dccp_sk(sk);
63 if (dp->dccps_options.dccpo_send_ack_vector)
64 dccp_ackvec_check_rcv_ackno(dp->dccps_hc_rx_ackvec, sk,
65 DCCP_SKB_CB(skb)->dccpd_ack_seq);
68 static int dccp_check_seqno(struct sock *sk, struct sk_buff *skb)
70 const struct dccp_hdr *dh = dccp_hdr(skb);
71 struct dccp_sock *dp = dccp_sk(sk);
75 * Step 5: Prepare sequence numbers for Sync
76 * If P.type == Sync or P.type == SyncAck,
77 * If S.AWL <= P.ackno <= S.AWH and P.seqno >= S.SWL,
78 * / * P is valid, so update sequence number variables
79 * accordingly. After this update, P will pass the tests
80 * in Step 6. A SyncAck is generated if necessary in
82 * Update S.GSR, S.SWL, S.SWH
84 * Drop packet and return
86 if (dh->dccph_type == DCCP_PKT_SYNC ||
87 dh->dccph_type == DCCP_PKT_SYNCACK) {
88 if (between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
89 dp->dccps_awl, dp->dccps_awh) &&
90 !before48(DCCP_SKB_CB(skb)->dccpd_seq, dp->dccps_swl))
91 dccp_update_gsr(sk, DCCP_SKB_CB(skb)->dccpd_seq);
97 * Step 6: Check sequence numbers
98 * Let LSWL = S.SWL and LAWL = S.AWL
99 * If P.type == CloseReq or P.type == Close or P.type == Reset,
100 * LSWL := S.GSR + 1, LAWL := S.GAR
101 * If LSWL <= P.seqno <= S.SWH
102 * and (P.ackno does not exist or LAWL <= P.ackno <= S.AWH),
103 * Update S.GSR, S.SWL, S.SWH
107 * Send Sync packet acknowledging P.seqno
108 * Drop packet and return
110 lswl = dp->dccps_swl;
111 lawl = dp->dccps_awl;
113 if (dh->dccph_type == DCCP_PKT_CLOSEREQ ||
114 dh->dccph_type == DCCP_PKT_CLOSE ||
115 dh->dccph_type == DCCP_PKT_RESET) {
116 lswl = dp->dccps_gsr;
117 dccp_inc_seqno(&lswl);
118 lawl = dp->dccps_gar;
121 if (between48(DCCP_SKB_CB(skb)->dccpd_seq, lswl, dp->dccps_swh) &&
122 (DCCP_SKB_CB(skb)->dccpd_ack_seq == DCCP_PKT_WITHOUT_ACK_SEQ ||
123 between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
124 lawl, dp->dccps_awh))) {
125 dccp_update_gsr(sk, DCCP_SKB_CB(skb)->dccpd_seq);
127 if (dh->dccph_type != DCCP_PKT_SYNC &&
128 (DCCP_SKB_CB(skb)->dccpd_ack_seq !=
129 DCCP_PKT_WITHOUT_ACK_SEQ))
130 dp->dccps_gar = DCCP_SKB_CB(skb)->dccpd_ack_seq;
132 LIMIT_NETDEBUG(KERN_WARNING "DCCP: Step 6 failed for %s packet, "
133 "(LSWL(%llu) <= P.seqno(%llu) <= S.SWH(%llu)) and "
134 "(P.ackno %s or LAWL(%llu) <= P.ackno(%llu) <= S.AWH(%llu), "
136 dccp_packet_name(dh->dccph_type),
137 (unsigned long long) lswl,
139 DCCP_SKB_CB(skb)->dccpd_seq,
140 (unsigned long long) dp->dccps_swh,
141 (DCCP_SKB_CB(skb)->dccpd_ack_seq ==
142 DCCP_PKT_WITHOUT_ACK_SEQ) ? "doesn't exist" : "exists",
143 (unsigned long long) lawl,
145 DCCP_SKB_CB(skb)->dccpd_ack_seq,
146 (unsigned long long) dp->dccps_awh);
147 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq, DCCP_PKT_SYNC);
154 static inline int __dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
155 const struct dccp_hdr *dh,
158 struct dccp_sock *dp = dccp_sk(sk);
160 switch (dccp_hdr(skb)->dccph_type) {
161 case DCCP_PKT_DATAACK:
164 * FIXME: check if sk_receive_queue is full, schedule DATA_DROPPED
167 __skb_pull(skb, dh->dccph_doff * 4);
168 __skb_queue_tail(&sk->sk_receive_queue, skb);
169 skb_set_owner_r(skb, sk);
170 sk->sk_data_ready(sk, 0);
176 * Step 9: Process Reset
177 * If P.type == Reset,
178 * Tear down connection
179 * S.state := TIMEWAIT
181 * Drop packet and return
184 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
186 case DCCP_PKT_CLOSEREQ:
187 dccp_rcv_closereq(sk, skb);
190 dccp_rcv_close(sk, skb);
192 case DCCP_PKT_REQUEST:
194 * or (S.is_server and P.type == Response)
195 * or (S.is_client and P.type == Request)
196 * or (S.state >= OPEN and P.type == Request
197 * and P.seqno >= S.OSR)
198 * or (S.state >= OPEN and P.type == Response
199 * and P.seqno >= S.OSR)
200 * or (S.state == RESPOND and P.type == Data),
201 * Send Sync packet acknowledging P.seqno
202 * Drop packet and return
204 if (dp->dccps_role != DCCP_ROLE_LISTEN)
207 case DCCP_PKT_RESPONSE:
208 if (dp->dccps_role != DCCP_ROLE_CLIENT)
211 if (!before48(DCCP_SKB_CB(skb)->dccpd_seq, dp->dccps_osr)) {
213 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
218 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
223 * As with DCCP-Ack packets, DCCP-Sync and DCCP-SyncAck packets
224 * MAY have non-zero-length application data areas, whose
225 * contents * receivers MUST ignore.
230 DCCP_INC_STATS_BH(DCCP_MIB_INERRS);
236 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
237 const struct dccp_hdr *dh, const unsigned len)
239 struct dccp_sock *dp = dccp_sk(sk);
241 if (dccp_check_seqno(sk, skb))
244 if (dccp_parse_options(sk, skb))
247 if (DCCP_SKB_CB(skb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
248 dccp_event_ack_recv(sk, skb);
250 if (dp->dccps_options.dccpo_send_ack_vector &&
251 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
252 DCCP_SKB_CB(skb)->dccpd_seq,
253 DCCP_ACKVEC_STATE_RECEIVED))
256 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
257 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
259 return __dccp_rcv_established(sk, skb, dh, len);
265 EXPORT_SYMBOL_GPL(dccp_rcv_established);
267 static int dccp_rcv_request_sent_state_process(struct sock *sk,
269 const struct dccp_hdr *dh,
273 * Step 4: Prepare sequence numbers in REQUEST
274 * If S.state == REQUEST,
275 * If (P.type == Response or P.type == Reset)
276 * and S.AWL <= P.ackno <= S.AWH,
277 * / * Set sequence number variables corresponding to the
278 * other endpoint, so P will pass the tests in Step 6 * /
279 * Set S.GSR, S.ISR, S.SWL, S.SWH
280 * / * Response processing continues in Step 10; Reset
281 * processing continues in Step 9 * /
283 if (dh->dccph_type == DCCP_PKT_RESPONSE) {
284 const struct inet_connection_sock *icsk = inet_csk(sk);
285 struct dccp_sock *dp = dccp_sk(sk);
287 /* Stop the REQUEST timer */
288 inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
289 BUG_TRAP(sk->sk_send_head != NULL);
290 __kfree_skb(sk->sk_send_head);
291 sk->sk_send_head = NULL;
293 if (!between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
294 dp->dccps_awl, dp->dccps_awh)) {
295 dccp_pr_debug("invalid ackno: S.AWL=%llu, "
296 "P.ackno=%llu, S.AWH=%llu \n",
297 (unsigned long long)dp->dccps_awl,
298 (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
299 (unsigned long long)dp->dccps_awh);
300 goto out_invalid_packet;
303 if (dccp_parse_options(sk, skb))
304 goto out_invalid_packet;
306 if (dp->dccps_options.dccpo_send_ack_vector &&
307 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
308 DCCP_SKB_CB(skb)->dccpd_seq,
309 DCCP_ACKVEC_STATE_RECEIVED))
310 goto out_invalid_packet; /* FIXME: change error code */
312 dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
313 dccp_update_gsr(sk, dp->dccps_isr);
315 * SWL and AWL are initially adjusted so that they are not less than
316 * the initial Sequence Numbers received and sent, respectively:
317 * SWL := max(GSR + 1 - floor(W/4), ISR),
318 * AWL := max(GSS - W' + 1, ISS).
319 * These adjustments MUST be applied only at the beginning of the
322 * AWL was adjusted in dccp_v4_connect -acme
324 dccp_set_seqno(&dp->dccps_swl,
325 max48(dp->dccps_swl, dp->dccps_isr));
327 dccp_sync_mss(sk, icsk->icsk_pmtu_cookie);
330 * Step 10: Process REQUEST state (second part)
331 * If S.state == REQUEST,
332 * / * If we get here, P is a valid Response from the
333 * server (see Step 4), and we should move to
334 * PARTOPEN state. PARTOPEN means send an Ack,
335 * don't send Data packets, retransmit Acks
336 * periodically, and always include any Init Cookie
337 * from the Response * /
338 * S.state := PARTOPEN
340 * Continue with S.state == PARTOPEN
341 * / * Step 12 will send the Ack completing the
342 * three-way handshake * /
344 dccp_set_state(sk, DCCP_PARTOPEN);
346 /* Make sure socket is routed, for correct metrics. */
347 icsk->icsk_af_ops->rebuild_header(sk);
349 if (!sock_flag(sk, SOCK_DEAD)) {
350 sk->sk_state_change(sk);
351 sk_wake_async(sk, 0, POLL_OUT);
354 if (sk->sk_write_pending || icsk->icsk_ack.pingpong ||
355 icsk->icsk_accept_queue.rskq_defer_accept) {
356 /* Save one ACK. Data will be ready after
357 * several ticks, if write_pending is set.
359 * It may be deleted, but with this feature tcpdumps
360 * look so _wonderfully_ clever, that I was not able
361 * to stand against the temptation 8) --ANK
364 * OK, in DCCP we can as well do a similar trick, its
365 * even in the draft, but there is no need for us to
366 * schedule an ack here, as dccp_sendmsg does this for
367 * us, also stated in the draft. -acme
377 /* dccp_v4_do_rcv will send a reset */
378 DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR;
382 static int dccp_rcv_respond_partopen_state_process(struct sock *sk,
384 const struct dccp_hdr *dh,
389 switch (dh->dccph_type) {
391 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
394 if (sk->sk_state == DCCP_RESPOND)
396 case DCCP_PKT_DATAACK:
399 * FIXME: we should be reseting the PARTOPEN (DELACK) timer
400 * here but only if we haven't used the DELACK timer for
401 * something else, like sending a delayed ack for a TIMESTAMP
402 * echo, etc, for now were not clearing it, sending an extra
403 * ACK when there is nothing else to do in DELACK is not a big
407 /* Stop the PARTOPEN timer */
408 if (sk->sk_state == DCCP_PARTOPEN)
409 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
411 dccp_sk(sk)->dccps_osr = DCCP_SKB_CB(skb)->dccpd_seq;
412 dccp_set_state(sk, DCCP_OPEN);
414 if (dh->dccph_type == DCCP_PKT_DATAACK ||
415 dh->dccph_type == DCCP_PKT_DATA) {
416 __dccp_rcv_established(sk, skb, dh, len);
417 queued = 1; /* packet was queued
418 (by __dccp_rcv_established) */
426 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
427 struct dccp_hdr *dh, unsigned len)
429 struct dccp_sock *dp = dccp_sk(sk);
430 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
431 const int old_state = sk->sk_state;
435 * Step 3: Process LISTEN state
436 * (Continuing from dccp_v4_do_rcv and dccp_v6_do_rcv)
438 * If S.state == LISTEN,
439 * If P.type == Request or P contains a valid Init Cookie
441 * * Must scan the packet's options to check for an Init
442 * Cookie. Only the Init Cookie is processed here,
443 * however; other options are processed in Step 8. This
444 * scan need only be performed if the endpoint uses Init
446 * * Generate a new socket and switch to that socket *
447 * Set S := new socket for this port pair
449 * Choose S.ISS (initial seqno) or set from Init Cookie
450 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
451 * Continue with S.state == RESPOND
452 * * A Response packet will be generated in Step 11 *
454 * Generate Reset(No Connection) unless P.type == Reset
455 * Drop packet and return
457 * NOTE: the check for the packet types is done in
458 * dccp_rcv_state_process
460 if (sk->sk_state == DCCP_LISTEN) {
461 if (dh->dccph_type == DCCP_PKT_REQUEST) {
462 if (inet_csk(sk)->icsk_af_ops->conn_request(sk,
466 /* FIXME: do congestion control initialization */
469 if (dh->dccph_type == DCCP_PKT_RESET)
472 /* Caller (dccp_v4_do_rcv) will send Reset */
473 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
477 if (sk->sk_state != DCCP_REQUESTING) {
478 if (dccp_check_seqno(sk, skb))
482 * Step 8: Process options and mark acknowledgeable
484 if (dccp_parse_options(sk, skb))
487 if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
488 dccp_event_ack_recv(sk, skb);
490 if (dp->dccps_options.dccpo_send_ack_vector &&
491 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
492 DCCP_SKB_CB(skb)->dccpd_seq,
493 DCCP_ACKVEC_STATE_RECEIVED))
496 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
497 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
501 * Step 9: Process Reset
502 * If P.type == Reset,
503 * Tear down connection
504 * S.state := TIMEWAIT
506 * Drop packet and return
508 if (dh->dccph_type == DCCP_PKT_RESET) {
510 * Queue the equivalent of TCP fin so that dccp_recvmsg
514 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
517 * Step 7: Check for unexpected packet types
518 * If (S.is_server and P.type == CloseReq)
519 * or (S.is_server and P.type == Response)
520 * or (S.is_client and P.type == Request)
521 * or (S.state == RESPOND and P.type == Data),
522 * Send Sync packet acknowledging P.seqno
523 * Drop packet and return
525 } else if ((dp->dccps_role != DCCP_ROLE_CLIENT &&
526 (dh->dccph_type == DCCP_PKT_RESPONSE ||
527 dh->dccph_type == DCCP_PKT_CLOSEREQ)) ||
528 (dp->dccps_role == DCCP_ROLE_CLIENT &&
529 dh->dccph_type == DCCP_PKT_REQUEST) ||
530 (sk->sk_state == DCCP_RESPOND &&
531 dh->dccph_type == DCCP_PKT_DATA)) {
532 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNC);
534 } else if (dh->dccph_type == DCCP_PKT_CLOSEREQ) {
535 dccp_rcv_closereq(sk, skb);
537 } else if (dh->dccph_type == DCCP_PKT_CLOSE) {
538 dccp_rcv_close(sk, skb);
542 if (unlikely(dh->dccph_type == DCCP_PKT_SYNC)) {
543 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNCACK);
547 switch (sk->sk_state) {
549 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
552 case DCCP_REQUESTING:
553 /* FIXME: do congestion control initialization */
555 queued = dccp_rcv_request_sent_state_process(sk, skb, dh, len);
564 queued = dccp_rcv_respond_partopen_state_process(sk, skb,
569 if (dh->dccph_type == DCCP_PKT_ACK ||
570 dh->dccph_type == DCCP_PKT_DATAACK) {
573 sk->sk_state_change(sk);
574 sk_wake_async(sk, 0, POLL_OUT);
586 EXPORT_SYMBOL_GPL(dccp_rcv_state_process);