1 /* 2 * net/dccp/input.c 3 * 4 * An implementation of the DCCP protocol 5 * Arnaldo Carvalho de Melo <acme@conectiva.com.br> 6 * 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. 11 */ 12 13 #include <linux/config.h> 14 #include <linux/dccp.h> 15 #include <linux/skbuff.h> 16 17 #include <net/sock.h> 18 19 #include "ackvec.h" 20 #include "ccid.h" 21 #include "dccp.h" 22 23 static void dccp_fin(struct sock *sk, struct sk_buff *skb) 24 { 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); 31 } 32 33 static void dccp_rcv_close(struct sock *sk, struct sk_buff *skb) 34 { 35 dccp_v4_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); 39 } 40 41 static void dccp_rcv_closereq(struct sock *sk, struct sk_buff *skb) 42 { 43 /* 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 48 */ 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; 52 } 53 54 if (sk->sk_state != DCCP_CLOSING) 55 dccp_set_state(sk, DCCP_CLOSING); 56 dccp_send_close(sk, 0); 57 } 58 59 static inline void dccp_event_ack_recv(struct sock *sk, struct sk_buff *skb) 60 { 61 struct dccp_sock *dp = dccp_sk(sk); 62 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); 66 } 67 68 static int dccp_check_seqno(struct sock *sk, struct sk_buff *skb) 69 { 70 const struct dccp_hdr *dh = dccp_hdr(skb); 71 struct dccp_sock *dp = dccp_sk(sk); 72 u64 lswl, lawl; 73 74 /* 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 85 */ 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; 94 } 95 96 /* 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 109 */ 110 lswl = dp->dccps_swl; 111 lawl = dp->dccps_awl; 112 113 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; 119 } 120 121 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); 126 127 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; 149 } 150 151 return 0; 152 } 153 154 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb, 155 const struct dccp_hdr *dh, const unsigned len) 156 { 157 struct dccp_sock *dp = dccp_sk(sk); 158 159 if (dccp_check_seqno(sk, skb)) 160 goto discard; 161 162 if (dccp_parse_options(sk, skb)) 163 goto discard; 164 165 if (DCCP_SKB_CB(skb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ) 166 dccp_event_ack_recv(sk, skb); 167 168 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)) 172 goto discard; 173 174 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); 176 177 switch (dccp_hdr(skb)->dccph_type) { 178 case DCCP_PKT_DATAACK: 179 case DCCP_PKT_DATA: 180 /* 181 * FIXME: check if sk_receive_queue is full, schedule DATA_DROPPED 182 * option if it is. 183 */ 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); 188 return 0; 189 case DCCP_PKT_ACK: 190 goto discard; 191 case DCCP_PKT_RESET: 192 /* 193 * Step 9: Process Reset 194 * If P.type == Reset, 195 * Tear down connection 196 * S.state := TIMEWAIT 197 * Set TIMEWAIT timer 198 * Drop packet and return 199 */ 200 dccp_fin(sk, skb); 201 dccp_time_wait(sk, DCCP_TIME_WAIT, 0); 202 return 0; 203 case DCCP_PKT_CLOSEREQ: 204 dccp_rcv_closereq(sk, skb); 205 goto discard; 206 case DCCP_PKT_CLOSE: 207 dccp_rcv_close(sk, skb); 208 return 0; 209 case DCCP_PKT_REQUEST: 210 /* Step 7 211 * 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 220 */ 221 if (dp->dccps_role != DCCP_ROLE_LISTEN) 222 goto send_sync; 223 goto check_seq; 224 case DCCP_PKT_RESPONSE: 225 if (dp->dccps_role != DCCP_ROLE_CLIENT) 226 goto send_sync; 227 check_seq: 228 if (!before48(DCCP_SKB_CB(skb)->dccpd_seq, dp->dccps_osr)) { 229 send_sync: 230 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq, 231 DCCP_PKT_SYNC); 232 } 233 break; 234 case DCCP_PKT_SYNC: 235 dccp_send_sync(sk, DCCP_SKB_CB(skb)->dccpd_seq, 236 DCCP_PKT_SYNCACK); 237 /* 238 * From the draft: 239 * 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. 243 */ 244 goto discard; 245 } 246 247 DCCP_INC_STATS_BH(DCCP_MIB_INERRS); 248 discard: 249 __kfree_skb(skb); 250 return 0; 251 } 252 253 static int dccp_rcv_request_sent_state_process(struct sock *sk, 254 struct sk_buff *skb, 255 const struct dccp_hdr *dh, 256 const unsigned len) 257 { 258 /* 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 * / 268 */ 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); 272 273 /* 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; 278 279 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; 287 } 288 289 dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq; 290 dccp_update_gsr(sk, dp->dccps_isr); 291 /* 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 297 * connection. 298 * 299 * AWL was adjusted in dccp_v4_connect -acme 300 */ 301 dccp_set_seqno(&dp->dccps_swl, 302 max48(dp->dccps_swl, dp->dccps_isr)); 303 304 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; 310 } 311 312 dccp_sync_mss(sk, dp->dccps_pmtu_cookie); 313 314 /* 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 324 * Set PARTOPEN timer 325 * Continue with S.state == PARTOPEN 326 * / * Step 12 will send the Ack completing the 327 * three-way handshake * / 328 */ 329 dccp_set_state(sk, DCCP_PARTOPEN); 330 331 /* Make sure socket is routed, for correct metrics. */ 332 inet_sk_rebuild_header(sk); 333 334 if (!sock_flag(sk, SOCK_DEAD)) { 335 sk->sk_state_change(sk); 336 sk_wake_async(sk, 0, POLL_OUT); 337 } 338 339 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. 343 * 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 347 */ 348 /* 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 353 */ 354 __kfree_skb(skb); 355 return 0; 356 } 357 dccp_send_ack(sk); 358 return -1; 359 } 360 361 out_invalid_packet: 362 /* dccp_v4_do_rcv will send a reset */ 363 DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR; 364 return 1; 365 } 366 367 static int dccp_rcv_respond_partopen_state_process(struct sock *sk, 368 struct sk_buff *skb, 369 const struct dccp_hdr *dh, 370 const unsigned len) 371 { 372 int queued = 0; 373 374 switch (dh->dccph_type) { 375 case DCCP_PKT_RESET: 376 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK); 377 break; 378 case DCCP_PKT_DATA: 379 if (sk->sk_state == DCCP_RESPOND) 380 break; 381 case DCCP_PKT_DATAACK: 382 case DCCP_PKT_ACK: 383 /* 384 * FIXME: we should be reseting the PARTOPEN (DELACK) timer 385 * here but only if we haven't used the DELACK timer for 386 * something else, like sending a delayed ack for a TIMESTAMP 387 * echo, etc, for now were not clearing it, sending an extra 388 * ACK when there is nothing else to do in DELACK is not a big 389 * deal after all. 390 */ 391 392 /* Stop the PARTOPEN timer */ 393 if (sk->sk_state == DCCP_PARTOPEN) 394 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK); 395 396 dccp_sk(sk)->dccps_osr = DCCP_SKB_CB(skb)->dccpd_seq; 397 dccp_set_state(sk, DCCP_OPEN); 398 399 if (dh->dccph_type == DCCP_PKT_DATAACK || 400 dh->dccph_type == DCCP_PKT_DATA) { 401 dccp_rcv_established(sk, skb, dh, len); 402 queued = 1; /* packet was queued 403 (by dccp_rcv_established) */ 404 } 405 break; 406 } 407 408 return queued; 409 } 410 411 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb, 412 struct dccp_hdr *dh, unsigned len) 413 { 414 struct dccp_sock *dp = dccp_sk(sk); 415 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); 416 const int old_state = sk->sk_state; 417 int queued = 0; 418 419 /* 420 * Step 3: Process LISTEN state 421 * (Continuing from dccp_v4_do_rcv and dccp_v6_do_rcv) 422 * 423 * If S.state == LISTEN, 424 * If P.type == Request or P contains a valid Init Cookie 425 * option, 426 * * Must scan the packet's options to check for an Init 427 * Cookie. Only the Init Cookie is processed here, 428 * however; other options are processed in Step 8. This 429 * scan need only be performed if the endpoint uses Init 430 * Cookies * 431 * * Generate a new socket and switch to that socket * 432 * Set S := new socket for this port pair 433 * S.state = RESPOND 434 * Choose S.ISS (initial seqno) or set from Init Cookie 435 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie 436 * Continue with S.state == RESPOND 437 * * A Response packet will be generated in Step 11 * 438 * Otherwise, 439 * Generate Reset(No Connection) unless P.type == Reset 440 * Drop packet and return 441 * 442 * NOTE: the check for the packet types is done in 443 * dccp_rcv_state_process 444 */ 445 if (sk->sk_state == DCCP_LISTEN) { 446 if (dh->dccph_type == DCCP_PKT_REQUEST) { 447 if (dccp_v4_conn_request(sk, skb) < 0) 448 return 1; 449 450 /* FIXME: do congestion control initialization */ 451 goto discard; 452 } 453 if (dh->dccph_type == DCCP_PKT_RESET) 454 goto discard; 455 456 /* Caller (dccp_v4_do_rcv) will send Reset */ 457 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION; 458 return 1; 459 } 460 461 if (sk->sk_state != DCCP_REQUESTING) { 462 if (dccp_check_seqno(sk, skb)) 463 goto discard; 464 465 /* 466 * Step 8: Process options and mark acknowledgeable 467 */ 468 if (dccp_parse_options(sk, skb)) 469 goto discard; 470 471 if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ) 472 dccp_event_ack_recv(sk, skb); 473 474 ccid_hc_rx_packet_recv(dp->dccps_hc_rx_ccid, sk, skb); 475 ccid_hc_tx_packet_recv(dp->dccps_hc_tx_ccid, sk, skb); 476 477 if (dp->dccps_options.dccpo_send_ack_vector && 478 dccp_ackvec_add(dp->dccps_hc_rx_ackvec, sk, 479 DCCP_SKB_CB(skb)->dccpd_seq, 480 DCCP_ACKVEC_STATE_RECEIVED)) 481 goto discard; 482 } 483 484 /* 485 * Step 9: Process Reset 486 * If P.type == Reset, 487 * Tear down connection 488 * S.state := TIMEWAIT 489 * Set TIMEWAIT timer 490 * Drop packet and return 491 */ 492 if (dh->dccph_type == DCCP_PKT_RESET) { 493 /* 494 * Queue the equivalent of TCP fin so that dccp_recvmsg 495 * exits the loop 496 */ 497 dccp_fin(sk, skb); 498 dccp_time_wait(sk, DCCP_TIME_WAIT, 0); 499 return 0; 500 /* 501 * Step 7: Check for unexpected packet types 502 * If (S.is_server and P.type == CloseReq) 503 * or (S.is_server and P.type == Response) 504 * or (S.is_client and P.type == Request) 505 * or (S.state == RESPOND and P.type == Data), 506 * Send Sync packet acknowledging P.seqno 507 * Drop packet and return 508 */ 509 } else if ((dp->dccps_role != DCCP_ROLE_CLIENT && 510 (dh->dccph_type == DCCP_PKT_RESPONSE || 511 dh->dccph_type == DCCP_PKT_CLOSEREQ)) || 512 (dp->dccps_role == DCCP_ROLE_CLIENT && 513 dh->dccph_type == DCCP_PKT_REQUEST) || 514 (sk->sk_state == DCCP_RESPOND && 515 dh->dccph_type == DCCP_PKT_DATA)) { 516 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNC); 517 goto discard; 518 } else if (dh->dccph_type == DCCP_PKT_CLOSEREQ) { 519 dccp_rcv_closereq(sk, skb); 520 goto discard; 521 } else if (dh->dccph_type == DCCP_PKT_CLOSE) { 522 dccp_rcv_close(sk, skb); 523 return 0; 524 } 525 526 if (unlikely(dh->dccph_type == DCCP_PKT_SYNC)) { 527 dccp_send_sync(sk, dcb->dccpd_seq, DCCP_PKT_SYNCACK); 528 goto discard; 529 } 530 531 switch (sk->sk_state) { 532 case DCCP_CLOSED: 533 dcb->dccpd_reset_code = DCCP_RESET_CODE_NO_CONNECTION; 534 return 1; 535 536 case DCCP_REQUESTING: 537 /* FIXME: do congestion control initialization */ 538 539 queued = dccp_rcv_request_sent_state_process(sk, skb, dh, len); 540 if (queued >= 0) 541 return queued; 542 543 __kfree_skb(skb); 544 return 0; 545 546 case DCCP_RESPOND: 547 case DCCP_PARTOPEN: 548 queued = dccp_rcv_respond_partopen_state_process(sk, skb, 549 dh, len); 550 break; 551 } 552 553 if (dh->dccph_type == DCCP_PKT_ACK || 554 dh->dccph_type == DCCP_PKT_DATAACK) { 555 switch (old_state) { 556 case DCCP_PARTOPEN: 557 sk->sk_state_change(sk); 558 sk_wake_async(sk, 0, POLL_OUT); 559 break; 560 } 561 } 562 563 if (!queued) { 564 discard: 565 __kfree_skb(skb); 566 } 567 return 0; 568 } 569