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 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
155 const struct dccp_hdr *dh, const unsigned len)
157 struct dccp_sock *dp = dccp_sk(sk);
159 if (dccp_check_seqno(sk, skb))
162 if (dccp_parse_options(sk, skb))
165 if (DCCP_SKB_CB(skb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
166 dccp_event_ack_recv(sk, skb);
168 if (dp->dccps_options.dccpo_send_ack_vector &&
169 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
170 DCCP_SKB_CB(skb)->dccpd_seq,
171 DCCP_ACKVEC_STATE_RECEIVED))
174 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
175 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
177 switch (dccp_hdr(skb)->dccph_type) {
178 case DCCP_PKT_DATAACK:
181 * FIXME: check if sk_receive_queue is full, schedule DATA_DROPPED
184 __skb_pull(skb, dh->dccph_doff * 4);
185 __skb_queue_tail(&sk->sk_receive_queue, skb);
186 skb_set_owner_r(skb, sk);
187 sk->sk_data_ready(sk, 0);
193 * Step 9: Process Reset
194 * If P.type == Reset,
195 * Tear down connection
196 * S.state := TIMEWAIT
198 * Drop packet and return
201 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
203 case DCCP_PKT_CLOSEREQ:
204 dccp_rcv_closereq(sk, skb);
207 dccp_rcv_close(sk, skb);
209 case DCCP_PKT_REQUEST:
211 * or (S.is_server and P.type == Response)
212 * or (S.is_client and P.type == Request)
213 * or (S.state >= OPEN and P.type == Request
214 * and P.seqno >= S.OSR)
215 * or (S.state >= OPEN and P.type == Response
216 * and P.seqno >= S.OSR)
217 * or (S.state == RESPOND and P.type == Data),
218 * Send Sync packet acknowledging P.seqno
219 * Drop packet and return
221 if (dp->dccps_role != DCCP_ROLE_LISTEN)
224 case DCCP_PKT_RESPONSE:
225 if (dp->dccps_role != DCCP_ROLE_CLIENT)
228 if (!before48(DCCP_SKB_CB(skb)->dccpd_seq, dp->dccps_osr)) {
230 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
235 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
240 * As with DCCP-Ack packets, DCCP-Sync and DCCP-SyncAck packets
241 * MAY have non-zero-length application data areas, whose
242 * contents * receivers MUST ignore.
247 DCCP_INC_STATS_BH(DCCP_MIB_INERRS);
253 static int dccp_rcv_request_sent_state_process(struct sock *sk,
255 const struct dccp_hdr *dh,
259 * Step 4: Prepare sequence numbers in REQUEST
260 * If S.state == REQUEST,
261 * If (P.type == Response or P.type == Reset)
262 * and S.AWL <= P.ackno <= S.AWH,
263 * / * Set sequence number variables corresponding to the
264 * other endpoint, so P will pass the tests in Step 6 * /
265 * Set S.GSR, S.ISR, S.SWL, S.SWH
266 * / * Response processing continues in Step 10; Reset
267 * processing continues in Step 9 * /
269 if (dh->dccph_type == DCCP_PKT_RESPONSE) {
270 const struct inet_connection_sock *icsk = inet_csk(sk);
271 struct dccp_sock *dp = dccp_sk(sk);
273 /* Stop the REQUEST timer */
274 inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
275 BUG_TRAP(sk->sk_send_head != NULL);
276 __kfree_skb(sk->sk_send_head);
277 sk->sk_send_head = NULL;
279 if (!between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
280 dp->dccps_awl, dp->dccps_awh)) {
281 dccp_pr_debug("invalid ackno: S.AWL=%llu, "
282 "P.ackno=%llu, S.AWH=%llu \n",
283 (unsigned long long)dp->dccps_awl,
284 (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
285 (unsigned long long)dp->dccps_awh);
286 goto out_invalid_packet;
289 dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
290 dccp_update_gsr(sk, dp->dccps_isr);
292 * SWL and AWL are initially adjusted so that they are not less than
293 * the initial Sequence Numbers received and sent, respectively:
294 * SWL := max(GSR + 1 - floor(W/4), ISR),
295 * AWL := max(GSS - W' + 1, ISS).
296 * These adjustments MUST be applied only at the beginning of the
299 * AWL was adjusted in dccp_v4_connect -acme
301 dccp_set_seqno(&dp->dccps_swl,
302 max48(dp->dccps_swl, dp->dccps_isr));
304 if (ccid_hc_rx_init(dp->dccps_hc_rx_ccid, sk) != 0 ||
305 ccid_hc_tx_init(dp->dccps_hc_tx_ccid, sk) != 0) {
306 ccid_hc_rx_exit(dp->dccps_hc_rx_ccid, sk);
307 ccid_hc_tx_exit(dp->dccps_hc_tx_ccid, sk);
308 /* FIXME: send appropriate RESET code */
309 goto out_invalid_packet;
312 dccp_sync_mss(sk, dp->dccps_pmtu_cookie);
315 * Step 10: Process REQUEST state (second part)
316 * If S.state == REQUEST,
317 * / * If we get here, P is a valid Response from the
318 * server (see Step 4), and we should move to
319 * PARTOPEN state. PARTOPEN means send an Ack,
320 * don't send Data packets, retransmit Acks
321 * periodically, and always include any Init Cookie
322 * from the Response * /
323 * S.state := PARTOPEN
325 * Continue with S.state == PARTOPEN
326 * / * Step 12 will send the Ack completing the
327 * three-way handshake * /
329 dccp_set_state(sk, DCCP_PARTOPEN);
331 /* Make sure socket is routed, for correct metrics. */
332 inet_sk_rebuild_header(sk);
334 if (!sock_flag(sk, SOCK_DEAD)) {
335 sk->sk_state_change(sk);
336 sk_wake_async(sk, 0, POLL_OUT);
339 if (sk->sk_write_pending || icsk->icsk_ack.pingpong ||
340 icsk->icsk_accept_queue.rskq_defer_accept) {
341 /* Save one ACK. Data will be ready after
342 * several ticks, if write_pending is set.
344 * It may be deleted, but with this feature tcpdumps
345 * look so _wonderfully_ clever, that I was not able
346 * to stand against the temptation 8) --ANK
349 * OK, in DCCP we can as well do a similar trick, its
350 * even in the draft, but there is no need for us to
351 * schedule an ack here, as dccp_sendmsg does this for
352 * us, also stated in the draft. -acme
362 /* dccp_v4_do_rcv will send a reset */
363 DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR;
367 static int dccp_rcv_respond_partopen_state_process(struct sock *sk,
369 const struct dccp_hdr *dh,
374 switch (dh->dccph_type) {
376 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
378 case DCCP_PKT_DATAACK:
381 * FIXME: we should be reseting the PARTOPEN (DELACK) timer
382 * here but only if we haven't used the DELACK timer for
383 * something else, like sending a delayed ack for a TIMESTAMP
384 * echo, etc, for now were not clearing it, sending an extra
385 * ACK when there is nothing else to do in DELACK is not a big
389 /* Stop the PARTOPEN timer */
390 if (sk->sk_state == DCCP_PARTOPEN)
391 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
393 dccp_sk(sk)->dccps_osr = DCCP_SKB_CB(skb)->dccpd_seq;
394 dccp_set_state(sk, DCCP_OPEN);
396 if (dh->dccph_type == DCCP_PKT_DATAACK) {
397 dccp_rcv_established(sk, skb, dh, len);
398 queued = 1; /* packet was queued
399 (by dccp_rcv_established) */
407 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
408 struct dccp_hdr *dh, unsigned len)
410 struct dccp_sock *dp = dccp_sk(sk);
411 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
412 const int old_state = sk->sk_state;
416 * Step 3: Process LISTEN state
417 * (Continuing from dccp_v4_do_rcv and dccp_v6_do_rcv)
419 * If S.state == LISTEN,
420 * If P.type == Request or P contains a valid Init Cookie
422 * * Must scan the packet's options to check for an Init
423 * Cookie. Only the Init Cookie is processed here,
424 * however; other options are processed in Step 8. This
425 * scan need only be performed if the endpoint uses Init
427 * * Generate a new socket and switch to that socket *
428 * Set S := new socket for this port pair
430 * Choose S.ISS (initial seqno) or set from Init Cookie
431 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
432 * Continue with S.state == RESPOND
433 * * A Response packet will be generated in Step 11 *
435 * Generate Reset(No Connection) unless P.type == Reset
436 * Drop packet and return
438 * NOTE: the check for the packet types is done in
439 * dccp_rcv_state_process
441 if (sk->sk_state == DCCP_LISTEN) {
442 if (dh->dccph_type == DCCP_PKT_REQUEST) {
443 if (dccp_v4_conn_request(sk, skb) < 0)
446 /* FIXME: do congestion control initialization */
449 if (dh->dccph_type == DCCP_PKT_RESET)
452 /* Caller (dccp_v4_do_rcv) will send Reset */
453 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
457 if (sk->sk_state != DCCP_REQUESTING) {
458 if (dccp_check_seqno(sk, skb))
462 * Step 8: Process options and mark acknowledgeable
464 if (dccp_parse_options(sk, skb))
467 if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
468 dccp_event_ack_recv(sk, skb);
470 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
471 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
473 if (dp->dccps_options.dccpo_send_ack_vector &&
474 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
475 DCCP_SKB_CB(skb)->dccpd_seq,
476 DCCP_ACKVEC_STATE_RECEIVED))
481 * Step 9: Process Reset
482 * If P.type == Reset,
483 * Tear down connection
484 * S.state := TIMEWAIT
486 * Drop packet and return
488 if (dh->dccph_type == DCCP_PKT_RESET) {
490 * Queue the equivalent of TCP fin so that dccp_recvmsg
494 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
497 * Step 7: Check for unexpected packet types
498 * If (S.is_server and P.type == CloseReq)
499 * or (S.is_server and P.type == Response)
500 * or (S.is_client and P.type == Request)
501 * or (S.state == RESPOND and P.type == Data),
502 * Send Sync packet acknowledging P.seqno
503 * Drop packet and return
505 } else if ((dp->dccps_role != DCCP_ROLE_CLIENT &&
506 (dh->dccph_type == DCCP_PKT_RESPONSE ||
507 dh->dccph_type == DCCP_PKT_CLOSEREQ)) ||
508 (dp->dccps_role == DCCP_ROLE_CLIENT &&
509 dh->dccph_type == DCCP_PKT_REQUEST) ||
510 (sk->sk_state == DCCP_RESPOND &&
511 dh->dccph_type == DCCP_PKT_DATA)) {
512 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNC);
514 } else if (dh->dccph_type == DCCP_PKT_CLOSEREQ) {
515 dccp_rcv_closereq(sk, skb);
517 } else if (dh->dccph_type == DCCP_PKT_CLOSE) {
518 dccp_rcv_close(sk, skb);
522 if (unlikely(dh->dccph_type == DCCP_PKT_SYNC)) {
523 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNCACK);
527 switch (sk->sk_state) {
529 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
532 case DCCP_REQUESTING:
533 /* FIXME: do congestion control initialization */
535 queued = dccp_rcv_request_sent_state_process(sk, skb, dh, len);
544 queued = dccp_rcv_respond_partopen_state_process(sk, skb,
549 if (dh->dccph_type == DCCP_PKT_ACK ||
550 dh->dccph_type == DCCP_PKT_DATAACK) {
553 sk->sk_state_change(sk);
554 sk_wake_async(sk, 0, POLL_OUT);