[NETFILTER]: Fix ip6tables breakage from {get,set}sockopt compat layer
[hh.org.git] / net / dccp / input.c
blobbfc53665516bf26a9d6066383c75ab21634fcd38
1 /*
2 * net/dccp/input.c
3 *
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>
17 #include <net/sock.h>
19 #include "ackvec.h"
20 #include "ccid.h"
21 #include "dccp.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_send_reset(sk, DCCP_RESET_CODE_CLOSED);
36 dccp_fin(sk, skb);
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);
51 return;
54 if (sk->sk_state != DCCP_CLOSING)
55 dccp_set_state(sk, DCCP_CLOSING);
56 dccp_send_close(sk, 0);
59 static void dccp_event_ack_recv(struct sock *sk, struct sk_buff *skb)
61 struct dccp_sock *dp = dccp_sk(sk);
63 if (dccp_msk(sk)->dccpms_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);
72 u64 lswl, lawl;
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
81 * Step 15 * /
82 * Update S.GSR, S.SWL, S.SWH
83 * Otherwise,
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);
92 else
93 return -1;
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
104 * If P.type != Sync,
105 * Update S.GAR
106 * Otherwise,
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;
131 } else {
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), "
135 "sending SYNC...\n",
136 dccp_packet_name(dh->dccph_type),
137 (unsigned long long) lswl,
138 (unsigned long long)
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,
144 (unsigned long long)
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);
148 return -1;
151 return 0;
154 static 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 switch (dccp_hdr(skb)->dccph_type) {
160 case DCCP_PKT_DATAACK:
161 case DCCP_PKT_DATA:
163 * FIXME: check if sk_receive_queue is full, schedule DATA_DROPPED
164 * option if it is.
166 __skb_pull(skb, dh->dccph_doff * 4);
167 __skb_queue_tail(&sk->sk_receive_queue, skb);
168 skb_set_owner_r(skb, sk);
169 sk->sk_data_ready(sk, 0);
170 return 0;
171 case DCCP_PKT_ACK:
172 goto discard;
173 case DCCP_PKT_RESET:
175 * Step 9: Process Reset
176 * If P.type == Reset,
177 * Tear down connection
178 * S.state := TIMEWAIT
179 * Set TIMEWAIT timer
180 * Drop packet and return
182 dccp_fin(sk, skb);
183 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
184 return 0;
185 case DCCP_PKT_CLOSEREQ:
186 dccp_rcv_closereq(sk, skb);
187 goto discard;
188 case DCCP_PKT_CLOSE:
189 dccp_rcv_close(sk, skb);
190 return 0;
191 case DCCP_PKT_REQUEST:
192 /* Step 7
193 * or (S.is_server and P.type == Response)
194 * or (S.is_client and P.type == Request)
195 * or (S.state >= OPEN and P.type == Request
196 * and P.seqno >= S.OSR)
197 * or (S.state >= OPEN and P.type == Response
198 * and P.seqno >= S.OSR)
199 * or (S.state == RESPOND and P.type == Data),
200 * Send Sync packet acknowledging P.seqno
201 * Drop packet and return
203 if (dp->dccps_role != DCCP_ROLE_LISTEN)
204 goto send_sync;
205 goto check_seq;
206 case DCCP_PKT_RESPONSE:
207 if (dp->dccps_role != DCCP_ROLE_CLIENT)
208 goto send_sync;
209 check_seq:
210 if (!before48(DCCP_SKB_CB(skb)->dccpd_seq, dp->dccps_osr)) {
211 send_sync:
212 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
213 DCCP_PKT_SYNC);
215 break;
216 case DCCP_PKT_SYNC:
217 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq,
218 DCCP_PKT_SYNCACK);
220 * From the draft:
222 * As with DCCP-Ack packets, DCCP-Sync and DCCP-SyncAck packets
223 * MAY have non-zero-length application data areas, whose
224 * contents * receivers MUST ignore.
226 goto discard;
229 DCCP_INC_STATS_BH(DCCP_MIB_INERRS);
230 discard:
231 __kfree_skb(skb);
232 return 0;
235 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
236 const struct dccp_hdr *dh, const unsigned len)
238 struct dccp_sock *dp = dccp_sk(sk);
240 if (dccp_check_seqno(sk, skb))
241 goto discard;
243 if (dccp_parse_options(sk, skb))
244 goto discard;
246 if (DCCP_SKB_CB(skb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
247 dccp_event_ack_recv(sk, skb);
249 if (dccp_msk(sk)->dccpms_send_ack_vector &&
250 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
251 DCCP_SKB_CB(skb)->dccpd_seq,
252 DCCP_ACKVEC_STATE_RECEIVED))
253 goto discard;
255 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
256 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
258 return __dccp_rcv_established(sk, skb, dh, len);
259 discard:
260 __kfree_skb(skb);
261 return 0;
264 EXPORT_SYMBOL_GPL(dccp_rcv_established);
266 static int dccp_rcv_request_sent_state_process(struct sock *sk,
267 struct sk_buff *skb,
268 const struct dccp_hdr *dh,
269 const unsigned len)
272 * Step 4: Prepare sequence numbers in REQUEST
273 * If S.state == REQUEST,
274 * If (P.type == Response or P.type == Reset)
275 * and S.AWL <= P.ackno <= S.AWH,
276 * / * Set sequence number variables corresponding to the
277 * other endpoint, so P will pass the tests in Step 6 * /
278 * Set S.GSR, S.ISR, S.SWL, S.SWH
279 * / * Response processing continues in Step 10; Reset
280 * processing continues in Step 9 * /
282 if (dh->dccph_type == DCCP_PKT_RESPONSE) {
283 const struct inet_connection_sock *icsk = inet_csk(sk);
284 struct dccp_sock *dp = dccp_sk(sk);
286 /* Stop the REQUEST timer */
287 inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
288 BUG_TRAP(sk->sk_send_head != NULL);
289 __kfree_skb(sk->sk_send_head);
290 sk->sk_send_head = NULL;
292 if (!between48(DCCP_SKB_CB(skb)->dccpd_ack_seq,
293 dp->dccps_awl, dp->dccps_awh)) {
294 dccp_pr_debug("invalid ackno: S.AWL=%llu, "
295 "P.ackno=%llu, S.AWH=%llu \n",
296 (unsigned long long)dp->dccps_awl,
297 (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
298 (unsigned long long)dp->dccps_awh);
299 goto out_invalid_packet;
302 if (dccp_parse_options(sk, skb))
303 goto out_invalid_packet;
305 if (dccp_msk(sk)->dccpms_send_ack_vector &&
306 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
307 DCCP_SKB_CB(skb)->dccpd_seq,
308 DCCP_ACKVEC_STATE_RECEIVED))
309 goto out_invalid_packet; /* FIXME: change error code */
311 dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
312 dccp_update_gsr(sk, dp->dccps_isr);
314 * SWL and AWL are initially adjusted so that they are not less than
315 * the initial Sequence Numbers received and sent, respectively:
316 * SWL := max(GSR + 1 - floor(W/4), ISR),
317 * AWL := max(GSS - W' + 1, ISS).
318 * These adjustments MUST be applied only at the beginning of the
319 * connection.
321 * AWL was adjusted in dccp_v4_connect -acme
323 dccp_set_seqno(&dp->dccps_swl,
324 max48(dp->dccps_swl, dp->dccps_isr));
326 dccp_sync_mss(sk, icsk->icsk_pmtu_cookie);
329 * Step 10: Process REQUEST state (second part)
330 * If S.state == REQUEST,
331 * / * If we get here, P is a valid Response from the
332 * server (see Step 4), and we should move to
333 * PARTOPEN state. PARTOPEN means send an Ack,
334 * don't send Data packets, retransmit Acks
335 * periodically, and always include any Init Cookie
336 * from the Response * /
337 * S.state := PARTOPEN
338 * Set PARTOPEN timer
339 * Continue with S.state == PARTOPEN
340 * / * Step 12 will send the Ack completing the
341 * three-way handshake * /
343 dccp_set_state(sk, DCCP_PARTOPEN);
345 /* Make sure socket is routed, for correct metrics. */
346 icsk->icsk_af_ops->rebuild_header(sk);
348 if (!sock_flag(sk, SOCK_DEAD)) {
349 sk->sk_state_change(sk);
350 sk_wake_async(sk, 0, POLL_OUT);
353 if (sk->sk_write_pending || icsk->icsk_ack.pingpong ||
354 icsk->icsk_accept_queue.rskq_defer_accept) {
355 /* Save one ACK. Data will be ready after
356 * several ticks, if write_pending is set.
358 * It may be deleted, but with this feature tcpdumps
359 * look so _wonderfully_ clever, that I was not able
360 * to stand against the temptation 8) --ANK
363 * OK, in DCCP we can as well do a similar trick, its
364 * even in the draft, but there is no need for us to
365 * schedule an ack here, as dccp_sendmsg does this for
366 * us, also stated in the draft. -acme
368 __kfree_skb(skb);
369 return 0;
371 dccp_send_ack(sk);
372 return -1;
375 out_invalid_packet:
376 /* dccp_v4_do_rcv will send a reset */
377 DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR;
378 return 1;
381 static int dccp_rcv_respond_partopen_state_process(struct sock *sk,
382 struct sk_buff *skb,
383 const struct dccp_hdr *dh,
384 const unsigned len)
386 int queued = 0;
388 switch (dh->dccph_type) {
389 case DCCP_PKT_RESET:
390 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
391 break;
392 case DCCP_PKT_DATA:
393 if (sk->sk_state == DCCP_RESPOND)
394 break;
395 case DCCP_PKT_DATAACK:
396 case DCCP_PKT_ACK:
398 * FIXME: we should be reseting the PARTOPEN (DELACK) timer
399 * here but only if we haven't used the DELACK timer for
400 * something else, like sending a delayed ack for a TIMESTAMP
401 * echo, etc, for now were not clearing it, sending an extra
402 * ACK when there is nothing else to do in DELACK is not a big
403 * deal after all.
406 /* Stop the PARTOPEN timer */
407 if (sk->sk_state == DCCP_PARTOPEN)
408 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
410 dccp_sk(sk)->dccps_osr = DCCP_SKB_CB(skb)->dccpd_seq;
411 dccp_set_state(sk, DCCP_OPEN);
413 if (dh->dccph_type == DCCP_PKT_DATAACK ||
414 dh->dccph_type == DCCP_PKT_DATA) {
415 __dccp_rcv_established(sk, skb, dh, len);
416 queued = 1; /* packet was queued
417 (by __dccp_rcv_established) */
419 break;
422 return queued;
425 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
426 struct dccp_hdr *dh, unsigned len)
428 struct dccp_sock *dp = dccp_sk(sk);
429 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
430 const int old_state = sk->sk_state;
431 int queued = 0;
434 * Step 3: Process LISTEN state
435 * (Continuing from dccp_v4_do_rcv and dccp_v6_do_rcv)
437 * If S.state == LISTEN,
438 * If P.type == Request or P contains a valid Init Cookie
439 * option,
440 * * Must scan the packet's options to check for an Init
441 * Cookie. Only the Init Cookie is processed here,
442 * however; other options are processed in Step 8. This
443 * scan need only be performed if the endpoint uses Init
444 * Cookies *
445 * * Generate a new socket and switch to that socket *
446 * Set S := new socket for this port pair
447 * S.state = RESPOND
448 * Choose S.ISS (initial seqno) or set from Init Cookie
449 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
450 * Continue with S.state == RESPOND
451 * * A Response packet will be generated in Step 11 *
452 * Otherwise,
453 * Generate Reset(No Connection) unless P.type == Reset
454 * Drop packet and return
456 * NOTE: the check for the packet types is done in
457 * dccp_rcv_state_process
459 if (sk->sk_state == DCCP_LISTEN) {
460 if (dh->dccph_type == DCCP_PKT_REQUEST) {
461 if (inet_csk(sk)->icsk_af_ops->conn_request(sk,
462 skb) < 0)
463 return 1;
465 /* FIXME: do congestion control initialization */
466 goto discard;
468 if (dh->dccph_type == DCCP_PKT_RESET)
469 goto discard;
471 /* Caller (dccp_v4_do_rcv) will send Reset */
472 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
473 return 1;
476 if (sk->sk_state != DCCP_REQUESTING) {
477 if (dccp_check_seqno(sk, skb))
478 goto discard;
481 * Step 8: Process options and mark acknowledgeable
483 if (dccp_parse_options(sk, skb))
484 goto discard;
486 if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
487 dccp_event_ack_recv(sk, skb);
489 if (dccp_msk(sk)->dccpms_send_ack_vector &&
490 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk,
491 DCCP_SKB_CB(skb)->dccpd_seq,
492 DCCP_ACKVEC_STATE_RECEIVED))
493 goto discard;
495 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb);
496 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb);
500 * Step 9: Process Reset
501 * If P.type == Reset,
502 * Tear down connection
503 * S.state := TIMEWAIT
504 * Set TIMEWAIT timer
505 * Drop packet and return
507 if (dh->dccph_type == DCCP_PKT_RESET) {
509 * Queue the equivalent of TCP fin so that dccp_recvmsg
510 * exits the loop
512 dccp_fin(sk, skb);
513 dccp_time_wait(sk, DCCP_TIME_WAIT, 0);
514 return 0;
516 * Step 7: Check for unexpected packet types
517 * If (S.is_server and P.type == CloseReq)
518 * or (S.is_server and P.type == Response)
519 * or (S.is_client and P.type == Request)
520 * or (S.state == RESPOND and P.type == Data),
521 * Send Sync packet acknowledging P.seqno
522 * Drop packet and return
524 } else if ((dp->dccps_role != DCCP_ROLE_CLIENT &&
525 (dh->dccph_type == DCCP_PKT_RESPONSE ||
526 dh->dccph_type == DCCP_PKT_CLOSEREQ)) ||
527 (dp->dccps_role == DCCP_ROLE_CLIENT &&
528 dh->dccph_type == DCCP_PKT_REQUEST) ||
529 (sk->sk_state == DCCP_RESPOND &&
530 dh->dccph_type == DCCP_PKT_DATA)) {
531 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNC);
532 goto discard;
533 } else if (dh->dccph_type == DCCP_PKT_CLOSEREQ) {
534 dccp_rcv_closereq(sk, skb);
535 goto discard;
536 } else if (dh->dccph_type == DCCP_PKT_CLOSE) {
537 dccp_rcv_close(sk, skb);
538 return 0;
541 if (unlikely(dh->dccph_type == DCCP_PKT_SYNC)) {
542 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNCACK);
543 goto discard;
546 switch (sk->sk_state) {
547 case DCCP_CLOSED:
548 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION;
549 return 1;
551 case DCCP_REQUESTING:
552 /* FIXME: do congestion control initialization */
554 queued = dccp_rcv_request_sent_state_process(sk, skb, dh, len);
555 if (queued >= 0)
556 return queued;
558 __kfree_skb(skb);
559 return 0;
561 case DCCP_RESPOND:
562 case DCCP_PARTOPEN:
563 queued = dccp_rcv_respond_partopen_state_process(sk, skb,
564 dh, len);
565 break;
568 if (dh->dccph_type == DCCP_PKT_ACK ||
569 dh->dccph_type == DCCP_PKT_DATAACK) {
570 switch (old_state) {
571 case DCCP_PARTOPEN:
572 sk->sk_state_change(sk);
573 sk_wake_async(sk, 0, POLL_OUT);
574 break;
578 if (!queued) {
579 discard:
580 __kfree_skb(skb);
582 return 0;
585 EXPORT_SYMBOL_GPL(dccp_rcv_state_process);