1 /* 2 * DCCP connection tracking protocol helper 3 * 4 * Copyright (c) 2005, 2006, 2008 Patrick McHardy <kaber@trash.net> 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License version 2 as 8 * published by the Free Software Foundation. 9 * 10 */ 11 #include <linux/kernel.h> 12 #include <linux/module.h> 13 #include <linux/init.h> 14 #include <linux/sysctl.h> 15 #include <linux/spinlock.h> 16 #include <linux/skbuff.h> 17 #include <linux/dccp.h> 18 #include <linux/slab.h> 19 20 #include <net/net_namespace.h> 21 #include <net/netns/generic.h> 22 23 #include <linux/netfilter/nfnetlink_conntrack.h> 24 #include <net/netfilter/nf_conntrack.h> 25 #include <net/netfilter/nf_conntrack_l4proto.h> 26 #include <net/netfilter/nf_conntrack_ecache.h> 27 #include <net/netfilter/nf_log.h> 28 29 /* Timeouts are based on values from RFC4340: 30 * 31 * - REQUEST: 32 * 33 * 8.1.2. Client Request 34 * 35 * A client MAY give up on its DCCP-Requests after some time 36 * (3 minutes, for example). 37 * 38 * - RESPOND: 39 * 40 * 8.1.3. Server Response 41 * 42 * It MAY also leave the RESPOND state for CLOSED after a timeout of 43 * not less than 4MSL (8 minutes); 44 * 45 * - PARTOPEN: 46 * 47 * 8.1.5. Handshake Completion 48 * 49 * If the client remains in PARTOPEN for more than 4MSL (8 minutes), 50 * it SHOULD reset the connection with Reset Code 2, "Aborted". 51 * 52 * - OPEN: 53 * 54 * The DCCP timestamp overflows after 11.9 hours. If the connection 55 * stays idle this long the sequence number won't be recognized 56 * as valid anymore. 57 * 58 * - CLOSEREQ/CLOSING: 59 * 60 * 8.3. Termination 61 * 62 * The retransmission timer should initially be set to go off in two 63 * round-trip times and should back off to not less than once every 64 * 64 seconds ... 65 * 66 * - TIMEWAIT: 67 * 68 * 4.3. States 69 * 70 * A server or client socket remains in this state for 2MSL (4 minutes) 71 * after the connection has been town down, ... 72 */ 73 74 #define DCCP_MSL (2 * 60 * HZ) 75 76 static const char * const dccp_state_names[] = { 77 [CT_DCCP_NONE] = "NONE", 78 [CT_DCCP_REQUEST] = "REQUEST", 79 [CT_DCCP_RESPOND] = "RESPOND", 80 [CT_DCCP_PARTOPEN] = "PARTOPEN", 81 [CT_DCCP_OPEN] = "OPEN", 82 [CT_DCCP_CLOSEREQ] = "CLOSEREQ", 83 [CT_DCCP_CLOSING] = "CLOSING", 84 [CT_DCCP_TIMEWAIT] = "TIMEWAIT", 85 [CT_DCCP_IGNORE] = "IGNORE", 86 [CT_DCCP_INVALID] = "INVALID", 87 }; 88 89 #define sNO CT_DCCP_NONE 90 #define sRQ CT_DCCP_REQUEST 91 #define sRS CT_DCCP_RESPOND 92 #define sPO CT_DCCP_PARTOPEN 93 #define sOP CT_DCCP_OPEN 94 #define sCR CT_DCCP_CLOSEREQ 95 #define sCG CT_DCCP_CLOSING 96 #define sTW CT_DCCP_TIMEWAIT 97 #define sIG CT_DCCP_IGNORE 98 #define sIV CT_DCCP_INVALID 99 100 /* 101 * DCCP state transition table 102 * 103 * The assumption is the same as for TCP tracking: 104 * 105 * We are the man in the middle. All the packets go through us but might 106 * get lost in transit to the destination. It is assumed that the destination 107 * can't receive segments we haven't seen. 108 * 109 * The following states exist: 110 * 111 * NONE: Initial state, expecting Request 112 * REQUEST: Request seen, waiting for Response from server 113 * RESPOND: Response from server seen, waiting for Ack from client 114 * PARTOPEN: Ack after Response seen, waiting for packet other than Response, 115 * Reset or Sync from server 116 * OPEN: Packet other than Response, Reset or Sync seen 117 * CLOSEREQ: CloseReq from server seen, expecting Close from client 118 * CLOSING: Close seen, expecting Reset 119 * TIMEWAIT: Reset seen 120 * IGNORE: Not determinable whether packet is valid 121 * 122 * Some states exist only on one side of the connection: REQUEST, RESPOND, 123 * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to 124 * the one it was in before. 125 * 126 * Packets are marked as ignored (sIG) if we don't know if they're valid 127 * (for example a reincarnation of a connection we didn't notice is dead 128 * already) and the server may send back a connection closing Reset or a 129 * Response. They're also used for Sync/SyncAck packets, which we don't 130 * care about. 131 */ 132 static const u_int8_t 133 dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = { 134 [CT_DCCP_ROLE_CLIENT] = { 135 [DCCP_PKT_REQUEST] = { 136 /* 137 * sNO -> sRQ Regular Request 138 * sRQ -> sRQ Retransmitted Request or reincarnation 139 * sRS -> sRS Retransmitted Request (apparently Response 140 * got lost after we saw it) or reincarnation 141 * sPO -> sIG Ignore, conntrack might be out of sync 142 * sOP -> sIG Ignore, conntrack might be out of sync 143 * sCR -> sIG Ignore, conntrack might be out of sync 144 * sCG -> sIG Ignore, conntrack might be out of sync 145 * sTW -> sRQ Reincarnation 146 * 147 * sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */ 148 sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ, 149 }, 150 [DCCP_PKT_RESPONSE] = { 151 /* 152 * sNO -> sIV Invalid 153 * sRQ -> sIG Ignore, might be response to ignored Request 154 * sRS -> sIG Ignore, might be response to ignored Request 155 * sPO -> sIG Ignore, might be response to ignored Request 156 * sOP -> sIG Ignore, might be response to ignored Request 157 * sCR -> sIG Ignore, might be response to ignored Request 158 * sCG -> sIG Ignore, might be response to ignored Request 159 * sTW -> sIV Invalid, reincarnation in reverse direction 160 * goes through sRQ 161 * 162 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 163 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV, 164 }, 165 [DCCP_PKT_ACK] = { 166 /* 167 * sNO -> sIV No connection 168 * sRQ -> sIV No connection 169 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.) 170 * sPO -> sPO Retransmitted Ack for Response, remain in PARTOPEN 171 * sOP -> sOP Regular ACK, remain in OPEN 172 * sCR -> sCR Ack in CLOSEREQ MAY be processed (8.3.) 173 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.) 174 * sTW -> sIV 175 * 176 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 177 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV 178 }, 179 [DCCP_PKT_DATA] = { 180 /* 181 * sNO -> sIV No connection 182 * sRQ -> sIV No connection 183 * sRS -> sIV No connection 184 * sPO -> sIV MUST use DataAck in PARTOPEN state (8.1.5.) 185 * sOP -> sOP Regular Data packet 186 * sCR -> sCR Data in CLOSEREQ MAY be processed (8.3.) 187 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 188 * sTW -> sIV 189 * 190 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 191 sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV, 192 }, 193 [DCCP_PKT_DATAACK] = { 194 /* 195 * sNO -> sIV No connection 196 * sRQ -> sIV No connection 197 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.) 198 * sPO -> sPO Remain in PARTOPEN state 199 * sOP -> sOP Regular DataAck packet in OPEN state 200 * sCR -> sCR DataAck in CLOSEREQ MAY be processed (8.3.) 201 * sCG -> sCG DataAck in CLOSING MAY be processed (8.3.) 202 * sTW -> sIV 203 * 204 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 205 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV 206 }, 207 [DCCP_PKT_CLOSEREQ] = { 208 /* 209 * CLOSEREQ may only be sent by the server. 210 * 211 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 212 sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV 213 }, 214 [DCCP_PKT_CLOSE] = { 215 /* 216 * sNO -> sIV No connection 217 * sRQ -> sIV No connection 218 * sRS -> sIV No connection 219 * sPO -> sCG Client-initiated close 220 * sOP -> sCG Client-initiated close 221 * sCR -> sCG Close in response to CloseReq (8.3.) 222 * sCG -> sCG Retransmit 223 * sTW -> sIV Late retransmit, already in TIME_WAIT 224 * 225 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 226 sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV 227 }, 228 [DCCP_PKT_RESET] = { 229 /* 230 * sNO -> sIV No connection 231 * sRQ -> sTW Sync received or timeout, SHOULD send Reset (8.1.1.) 232 * sRS -> sTW Response received without Request 233 * sPO -> sTW Timeout, SHOULD send Reset (8.1.5.) 234 * sOP -> sTW Connection reset 235 * sCR -> sTW Connection reset 236 * sCG -> sTW Connection reset 237 * sTW -> sIG Ignore (don't refresh timer) 238 * 239 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 240 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG 241 }, 242 [DCCP_PKT_SYNC] = { 243 /* 244 * We currently ignore Sync packets 245 * 246 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 247 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 248 }, 249 [DCCP_PKT_SYNCACK] = { 250 /* 251 * We currently ignore SyncAck packets 252 * 253 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 254 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 255 }, 256 }, 257 [CT_DCCP_ROLE_SERVER] = { 258 [DCCP_PKT_REQUEST] = { 259 /* 260 * sNO -> sIV Invalid 261 * sRQ -> sIG Ignore, conntrack might be out of sync 262 * sRS -> sIG Ignore, conntrack might be out of sync 263 * sPO -> sIG Ignore, conntrack might be out of sync 264 * sOP -> sIG Ignore, conntrack might be out of sync 265 * sCR -> sIG Ignore, conntrack might be out of sync 266 * sCG -> sIG Ignore, conntrack might be out of sync 267 * sTW -> sRQ Reincarnation, must reverse roles 268 * 269 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 270 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ 271 }, 272 [DCCP_PKT_RESPONSE] = { 273 /* 274 * sNO -> sIV Response without Request 275 * sRQ -> sRS Response to clients Request 276 * sRS -> sRS Retransmitted Response (8.1.3. SHOULD NOT) 277 * sPO -> sIG Response to an ignored Request or late retransmit 278 * sOP -> sIG Ignore, might be response to ignored Request 279 * sCR -> sIG Ignore, might be response to ignored Request 280 * sCG -> sIG Ignore, might be response to ignored Request 281 * sTW -> sIV Invalid, Request from client in sTW moves to sRQ 282 * 283 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 284 sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV 285 }, 286 [DCCP_PKT_ACK] = { 287 /* 288 * sNO -> sIV No connection 289 * sRQ -> sIV No connection 290 * sRS -> sIV No connection 291 * sPO -> sOP Enter OPEN state (8.1.5.) 292 * sOP -> sOP Regular Ack in OPEN state 293 * sCR -> sIV Waiting for Close from client 294 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.) 295 * sTW -> sIV 296 * 297 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 298 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 299 }, 300 [DCCP_PKT_DATA] = { 301 /* 302 * sNO -> sIV No connection 303 * sRQ -> sIV No connection 304 * sRS -> sIV No connection 305 * sPO -> sOP Enter OPEN state (8.1.5.) 306 * sOP -> sOP Regular Data packet in OPEN state 307 * sCR -> sIV Waiting for Close from client 308 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 309 * sTW -> sIV 310 * 311 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 312 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 313 }, 314 [DCCP_PKT_DATAACK] = { 315 /* 316 * sNO -> sIV No connection 317 * sRQ -> sIV No connection 318 * sRS -> sIV No connection 319 * sPO -> sOP Enter OPEN state (8.1.5.) 320 * sOP -> sOP Regular DataAck in OPEN state 321 * sCR -> sIV Waiting for Close from client 322 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 323 * sTW -> sIV 324 * 325 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 326 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 327 }, 328 [DCCP_PKT_CLOSEREQ] = { 329 /* 330 * sNO -> sIV No connection 331 * sRQ -> sIV No connection 332 * sRS -> sIV No connection 333 * sPO -> sOP -> sCR Move directly to CLOSEREQ (8.1.5.) 334 * sOP -> sCR CloseReq in OPEN state 335 * sCR -> sCR Retransmit 336 * sCG -> sCR Simultaneous close, client sends another Close 337 * sTW -> sIV Already closed 338 * 339 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 340 sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV 341 }, 342 [DCCP_PKT_CLOSE] = { 343 /* 344 * sNO -> sIV No connection 345 * sRQ -> sIV No connection 346 * sRS -> sIV No connection 347 * sPO -> sOP -> sCG Move direcly to CLOSING 348 * sOP -> sCG Move to CLOSING 349 * sCR -> sIV Close after CloseReq is invalid 350 * sCG -> sCG Retransmit 351 * sTW -> sIV Already closed 352 * 353 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 354 sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV 355 }, 356 [DCCP_PKT_RESET] = { 357 /* 358 * sNO -> sIV No connection 359 * sRQ -> sTW Reset in response to Request 360 * sRS -> sTW Timeout, SHOULD send Reset (8.1.3.) 361 * sPO -> sTW Timeout, SHOULD send Reset (8.1.3.) 362 * sOP -> sTW 363 * sCR -> sTW 364 * sCG -> sTW 365 * sTW -> sIG Ignore (don't refresh timer) 366 * 367 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */ 368 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG 369 }, 370 [DCCP_PKT_SYNC] = { 371 /* 372 * We currently ignore Sync packets 373 * 374 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 375 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 376 }, 377 [DCCP_PKT_SYNCACK] = { 378 /* 379 * We currently ignore SyncAck packets 380 * 381 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 382 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 383 }, 384 }, 385 }; 386 387 /* this module per-net specifics */ 388 static int dccp_net_id __read_mostly; 389 struct dccp_net { 390 int dccp_loose; 391 unsigned int dccp_timeout[CT_DCCP_MAX + 1]; 392 #ifdef CONFIG_SYSCTL 393 struct ctl_table_header *sysctl_header; 394 struct ctl_table *sysctl_table; 395 #endif 396 }; 397 398 static inline struct dccp_net *dccp_pernet(struct net *net) 399 { 400 return net_generic(net, dccp_net_id); 401 } 402 403 static bool dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff, 404 struct nf_conntrack_tuple *tuple) 405 { 406 struct dccp_hdr _hdr, *dh; 407 408 dh = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr); 409 if (dh == NULL) 410 return false; 411 412 tuple->src.u.dccp.port = dh->dccph_sport; 413 tuple->dst.u.dccp.port = dh->dccph_dport; 414 return true; 415 } 416 417 static bool dccp_invert_tuple(struct nf_conntrack_tuple *inv, 418 const struct nf_conntrack_tuple *tuple) 419 { 420 inv->src.u.dccp.port = tuple->dst.u.dccp.port; 421 inv->dst.u.dccp.port = tuple->src.u.dccp.port; 422 return true; 423 } 424 425 static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb, 426 unsigned int dataoff, unsigned int *timeouts) 427 { 428 struct net *net = nf_ct_net(ct); 429 struct dccp_net *dn; 430 struct dccp_hdr _dh, *dh; 431 const char *msg; 432 u_int8_t state; 433 434 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh); 435 BUG_ON(dh == NULL); 436 437 state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE]; 438 switch (state) { 439 default: 440 dn = dccp_pernet(net); 441 if (dn->dccp_loose == 0) { 442 msg = "nf_ct_dccp: not picking up existing connection "; 443 goto out_invalid; 444 } 445 case CT_DCCP_REQUEST: 446 break; 447 case CT_DCCP_INVALID: 448 msg = "nf_ct_dccp: invalid state transition "; 449 goto out_invalid; 450 } 451 452 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT; 453 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER; 454 ct->proto.dccp.state = CT_DCCP_NONE; 455 ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST; 456 ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL; 457 ct->proto.dccp.handshake_seq = 0; 458 return true; 459 460 out_invalid: 461 if (LOG_INVALID(net, IPPROTO_DCCP)) 462 nf_log_packet(nf_ct_l3num(ct), 0, skb, NULL, NULL, NULL, msg); 463 return false; 464 } 465 466 static u64 dccp_ack_seq(const struct dccp_hdr *dh) 467 { 468 const struct dccp_hdr_ack_bits *dhack; 469 470 dhack = (void *)dh + __dccp_basic_hdr_len(dh); 471 return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) + 472 ntohl(dhack->dccph_ack_nr_low); 473 } 474 475 static unsigned int *dccp_get_timeouts(struct net *net) 476 { 477 return dccp_pernet(net)->dccp_timeout; 478 } 479 480 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb, 481 unsigned int dataoff, enum ip_conntrack_info ctinfo, 482 u_int8_t pf, unsigned int hooknum, 483 unsigned int *timeouts) 484 { 485 struct net *net = nf_ct_net(ct); 486 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); 487 struct dccp_hdr _dh, *dh; 488 u_int8_t type, old_state, new_state; 489 enum ct_dccp_roles role; 490 491 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh); 492 BUG_ON(dh == NULL); 493 type = dh->dccph_type; 494 495 if (type == DCCP_PKT_RESET && 496 !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { 497 /* Tear down connection immediately if only reply is a RESET */ 498 nf_ct_kill_acct(ct, ctinfo, skb); 499 return NF_ACCEPT; 500 } 501 502 spin_lock_bh(&ct->lock); 503 504 role = ct->proto.dccp.role[dir]; 505 old_state = ct->proto.dccp.state; 506 new_state = dccp_state_table[role][type][old_state]; 507 508 switch (new_state) { 509 case CT_DCCP_REQUEST: 510 if (old_state == CT_DCCP_TIMEWAIT && 511 role == CT_DCCP_ROLE_SERVER) { 512 /* Reincarnation in the reverse direction: reopen and 513 * reverse client/server roles. */ 514 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT; 515 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER; 516 } 517 break; 518 case CT_DCCP_RESPOND: 519 if (old_state == CT_DCCP_REQUEST) 520 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh); 521 break; 522 case CT_DCCP_PARTOPEN: 523 if (old_state == CT_DCCP_RESPOND && 524 type == DCCP_PKT_ACK && 525 dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq) 526 set_bit(IPS_ASSURED_BIT, &ct->status); 527 break; 528 case CT_DCCP_IGNORE: 529 /* 530 * Connection tracking might be out of sync, so we ignore 531 * packets that might establish a new connection and resync 532 * if the server responds with a valid Response. 533 */ 534 if (ct->proto.dccp.last_dir == !dir && 535 ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST && 536 type == DCCP_PKT_RESPONSE) { 537 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT; 538 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER; 539 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh); 540 new_state = CT_DCCP_RESPOND; 541 break; 542 } 543 ct->proto.dccp.last_dir = dir; 544 ct->proto.dccp.last_pkt = type; 545 546 spin_unlock_bh(&ct->lock); 547 if (LOG_INVALID(net, IPPROTO_DCCP)) 548 nf_log_packet(pf, 0, skb, NULL, NULL, NULL, 549 "nf_ct_dccp: invalid packet ignored "); 550 return NF_ACCEPT; 551 case CT_DCCP_INVALID: 552 spin_unlock_bh(&ct->lock); 553 if (LOG_INVALID(net, IPPROTO_DCCP)) 554 nf_log_packet(pf, 0, skb, NULL, NULL, NULL, 555 "nf_ct_dccp: invalid state transition "); 556 return -NF_ACCEPT; 557 } 558 559 ct->proto.dccp.last_dir = dir; 560 ct->proto.dccp.last_pkt = type; 561 ct->proto.dccp.state = new_state; 562 spin_unlock_bh(&ct->lock); 563 564 if (new_state != old_state) 565 nf_conntrack_event_cache(IPCT_PROTOINFO, ct); 566 567 nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]); 568 569 return NF_ACCEPT; 570 } 571 572 static int dccp_error(struct net *net, struct nf_conn *tmpl, 573 struct sk_buff *skb, unsigned int dataoff, 574 enum ip_conntrack_info *ctinfo, 575 u_int8_t pf, unsigned int hooknum) 576 { 577 struct dccp_hdr _dh, *dh; 578 unsigned int dccp_len = skb->len - dataoff; 579 unsigned int cscov; 580 const char *msg; 581 582 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh); 583 if (dh == NULL) { 584 msg = "nf_ct_dccp: short packet "; 585 goto out_invalid; 586 } 587 588 if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) || 589 dh->dccph_doff * 4 > dccp_len) { 590 msg = "nf_ct_dccp: truncated/malformed packet "; 591 goto out_invalid; 592 } 593 594 cscov = dccp_len; 595 if (dh->dccph_cscov) { 596 cscov = (dh->dccph_cscov - 1) * 4; 597 if (cscov > dccp_len) { 598 msg = "nf_ct_dccp: bad checksum coverage "; 599 goto out_invalid; 600 } 601 } 602 603 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING && 604 nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP, 605 pf)) { 606 msg = "nf_ct_dccp: bad checksum "; 607 goto out_invalid; 608 } 609 610 if (dh->dccph_type >= DCCP_PKT_INVALID) { 611 msg = "nf_ct_dccp: reserved packet type "; 612 goto out_invalid; 613 } 614 615 return NF_ACCEPT; 616 617 out_invalid: 618 if (LOG_INVALID(net, IPPROTO_DCCP)) 619 nf_log_packet(pf, 0, skb, NULL, NULL, NULL, msg); 620 return -NF_ACCEPT; 621 } 622 623 static int dccp_print_tuple(struct seq_file *s, 624 const struct nf_conntrack_tuple *tuple) 625 { 626 return seq_printf(s, "sport=%hu dport=%hu ", 627 ntohs(tuple->src.u.dccp.port), 628 ntohs(tuple->dst.u.dccp.port)); 629 } 630 631 static int dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct) 632 { 633 return seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]); 634 } 635 636 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 637 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla, 638 struct nf_conn *ct) 639 { 640 struct nlattr *nest_parms; 641 642 spin_lock_bh(&ct->lock); 643 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED); 644 if (!nest_parms) 645 goto nla_put_failure; 646 if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) || 647 nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE, 648 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) || 649 nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ, 650 cpu_to_be64(ct->proto.dccp.handshake_seq))) 651 goto nla_put_failure; 652 nla_nest_end(skb, nest_parms); 653 spin_unlock_bh(&ct->lock); 654 return 0; 655 656 nla_put_failure: 657 spin_unlock_bh(&ct->lock); 658 return -1; 659 } 660 661 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = { 662 [CTA_PROTOINFO_DCCP_STATE] = { .type = NLA_U8 }, 663 [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 }, 664 [CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 }, 665 }; 666 667 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct) 668 { 669 struct nlattr *attr = cda[CTA_PROTOINFO_DCCP]; 670 struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1]; 671 int err; 672 673 if (!attr) 674 return 0; 675 676 err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr, 677 dccp_nla_policy); 678 if (err < 0) 679 return err; 680 681 if (!tb[CTA_PROTOINFO_DCCP_STATE] || 682 !tb[CTA_PROTOINFO_DCCP_ROLE] || 683 nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX || 684 nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) { 685 return -EINVAL; 686 } 687 688 spin_lock_bh(&ct->lock); 689 ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]); 690 if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) { 691 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT; 692 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER; 693 } else { 694 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER; 695 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT; 696 } 697 if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) { 698 ct->proto.dccp.handshake_seq = 699 be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ])); 700 } 701 spin_unlock_bh(&ct->lock); 702 return 0; 703 } 704 705 static int dccp_nlattr_size(void) 706 { 707 return nla_total_size(0) /* CTA_PROTOINFO_DCCP */ 708 + nla_policy_len(dccp_nla_policy, CTA_PROTOINFO_DCCP_MAX + 1); 709 } 710 711 #endif 712 713 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 714 715 #include <linux/netfilter/nfnetlink.h> 716 #include <linux/netfilter/nfnetlink_cttimeout.h> 717 718 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[], void *data) 719 { 720 struct dccp_net *dn = dccp_pernet(&init_net); 721 unsigned int *timeouts = data; 722 int i; 723 724 /* set default DCCP timeouts. */ 725 for (i=0; i<CT_DCCP_MAX; i++) 726 timeouts[i] = dn->dccp_timeout[i]; 727 728 /* there's a 1:1 mapping between attributes and protocol states. */ 729 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) { 730 if (tb[i]) { 731 timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ; 732 } 733 } 734 return 0; 735 } 736 737 static int 738 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data) 739 { 740 const unsigned int *timeouts = data; 741 int i; 742 743 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) { 744 if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ))) 745 goto nla_put_failure; 746 } 747 return 0; 748 749 nla_put_failure: 750 return -ENOSPC; 751 } 752 753 static const struct nla_policy 754 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = { 755 [CTA_TIMEOUT_DCCP_REQUEST] = { .type = NLA_U32 }, 756 [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 }, 757 [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 }, 758 [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 }, 759 [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 }, 760 [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 }, 761 [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 }, 762 }; 763 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ 764 765 #ifdef CONFIG_SYSCTL 766 /* template, data assigned later */ 767 static struct ctl_table dccp_sysctl_table[] = { 768 { 769 .procname = "nf_conntrack_dccp_timeout_request", 770 .maxlen = sizeof(unsigned int), 771 .mode = 0644, 772 .proc_handler = proc_dointvec_jiffies, 773 }, 774 { 775 .procname = "nf_conntrack_dccp_timeout_respond", 776 .maxlen = sizeof(unsigned int), 777 .mode = 0644, 778 .proc_handler = proc_dointvec_jiffies, 779 }, 780 { 781 .procname = "nf_conntrack_dccp_timeout_partopen", 782 .maxlen = sizeof(unsigned int), 783 .mode = 0644, 784 .proc_handler = proc_dointvec_jiffies, 785 }, 786 { 787 .procname = "nf_conntrack_dccp_timeout_open", 788 .maxlen = sizeof(unsigned int), 789 .mode = 0644, 790 .proc_handler = proc_dointvec_jiffies, 791 }, 792 { 793 .procname = "nf_conntrack_dccp_timeout_closereq", 794 .maxlen = sizeof(unsigned int), 795 .mode = 0644, 796 .proc_handler = proc_dointvec_jiffies, 797 }, 798 { 799 .procname = "nf_conntrack_dccp_timeout_closing", 800 .maxlen = sizeof(unsigned int), 801 .mode = 0644, 802 .proc_handler = proc_dointvec_jiffies, 803 }, 804 { 805 .procname = "nf_conntrack_dccp_timeout_timewait", 806 .maxlen = sizeof(unsigned int), 807 .mode = 0644, 808 .proc_handler = proc_dointvec_jiffies, 809 }, 810 { 811 .procname = "nf_conntrack_dccp_loose", 812 .maxlen = sizeof(int), 813 .mode = 0644, 814 .proc_handler = proc_dointvec, 815 }, 816 { } 817 }; 818 #endif /* CONFIG_SYSCTL */ 819 820 static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = { 821 .l3proto = AF_INET, 822 .l4proto = IPPROTO_DCCP, 823 .name = "dccp", 824 .pkt_to_tuple = dccp_pkt_to_tuple, 825 .invert_tuple = dccp_invert_tuple, 826 .new = dccp_new, 827 .packet = dccp_packet, 828 .get_timeouts = dccp_get_timeouts, 829 .error = dccp_error, 830 .print_tuple = dccp_print_tuple, 831 .print_conntrack = dccp_print_conntrack, 832 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 833 .to_nlattr = dccp_to_nlattr, 834 .nlattr_size = dccp_nlattr_size, 835 .from_nlattr = nlattr_to_dccp, 836 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, 837 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size, 838 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, 839 .nla_policy = nf_ct_port_nla_policy, 840 #endif 841 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 842 .ctnl_timeout = { 843 .nlattr_to_obj = dccp_timeout_nlattr_to_obj, 844 .obj_to_nlattr = dccp_timeout_obj_to_nlattr, 845 .nlattr_max = CTA_TIMEOUT_DCCP_MAX, 846 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX, 847 .nla_policy = dccp_timeout_nla_policy, 848 }, 849 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ 850 }; 851 852 static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = { 853 .l3proto = AF_INET6, 854 .l4proto = IPPROTO_DCCP, 855 .name = "dccp", 856 .pkt_to_tuple = dccp_pkt_to_tuple, 857 .invert_tuple = dccp_invert_tuple, 858 .new = dccp_new, 859 .packet = dccp_packet, 860 .get_timeouts = dccp_get_timeouts, 861 .error = dccp_error, 862 .print_tuple = dccp_print_tuple, 863 .print_conntrack = dccp_print_conntrack, 864 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 865 .to_nlattr = dccp_to_nlattr, 866 .nlattr_size = dccp_nlattr_size, 867 .from_nlattr = nlattr_to_dccp, 868 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, 869 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size, 870 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, 871 .nla_policy = nf_ct_port_nla_policy, 872 #endif 873 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) 874 .ctnl_timeout = { 875 .nlattr_to_obj = dccp_timeout_nlattr_to_obj, 876 .obj_to_nlattr = dccp_timeout_obj_to_nlattr, 877 .nlattr_max = CTA_TIMEOUT_DCCP_MAX, 878 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX, 879 .nla_policy = dccp_timeout_nla_policy, 880 }, 881 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ 882 }; 883 884 static __net_init int dccp_net_init(struct net *net) 885 { 886 struct dccp_net *dn = dccp_pernet(net); 887 888 /* default values */ 889 dn->dccp_loose = 1; 890 dn->dccp_timeout[CT_DCCP_REQUEST] = 2 * DCCP_MSL; 891 dn->dccp_timeout[CT_DCCP_RESPOND] = 4 * DCCP_MSL; 892 dn->dccp_timeout[CT_DCCP_PARTOPEN] = 4 * DCCP_MSL; 893 dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ; 894 dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ; 895 dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ; 896 dn->dccp_timeout[CT_DCCP_TIMEWAIT] = 2 * DCCP_MSL; 897 898 #ifdef CONFIG_SYSCTL 899 dn->sysctl_table = kmemdup(dccp_sysctl_table, 900 sizeof(dccp_sysctl_table), GFP_KERNEL); 901 if (!dn->sysctl_table) 902 return -ENOMEM; 903 904 dn->sysctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST]; 905 dn->sysctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND]; 906 dn->sysctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN]; 907 dn->sysctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN]; 908 dn->sysctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ]; 909 dn->sysctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING]; 910 dn->sysctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT]; 911 dn->sysctl_table[7].data = &dn->dccp_loose; 912 913 dn->sysctl_header = register_net_sysctl_table(net, 914 nf_net_netfilter_sysctl_path, dn->sysctl_table); 915 if (!dn->sysctl_header) { 916 kfree(dn->sysctl_table); 917 return -ENOMEM; 918 } 919 #endif 920 921 return 0; 922 } 923 924 static __net_exit void dccp_net_exit(struct net *net) 925 { 926 struct dccp_net *dn = dccp_pernet(net); 927 #ifdef CONFIG_SYSCTL 928 unregister_net_sysctl_table(dn->sysctl_header); 929 kfree(dn->sysctl_table); 930 #endif 931 } 932 933 static struct pernet_operations dccp_net_ops = { 934 .init = dccp_net_init, 935 .exit = dccp_net_exit, 936 .id = &dccp_net_id, 937 .size = sizeof(struct dccp_net), 938 }; 939 940 static int __init nf_conntrack_proto_dccp_init(void) 941 { 942 int err; 943 944 err = register_pernet_subsys(&dccp_net_ops); 945 if (err < 0) 946 goto err1; 947 948 err = nf_conntrack_l4proto_register(&dccp_proto4); 949 if (err < 0) 950 goto err2; 951 952 err = nf_conntrack_l4proto_register(&dccp_proto6); 953 if (err < 0) 954 goto err3; 955 return 0; 956 957 err3: 958 nf_conntrack_l4proto_unregister(&dccp_proto4); 959 err2: 960 unregister_pernet_subsys(&dccp_net_ops); 961 err1: 962 return err; 963 } 964 965 static void __exit nf_conntrack_proto_dccp_fini(void) 966 { 967 unregister_pernet_subsys(&dccp_net_ops); 968 nf_conntrack_l4proto_unregister(&dccp_proto6); 969 nf_conntrack_l4proto_unregister(&dccp_proto4); 970 } 971 972 module_init(nf_conntrack_proto_dccp_init); 973 module_exit(nf_conntrack_proto_dccp_fini); 974 975 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 976 MODULE_DESCRIPTION("DCCP connection tracking protocol helper"); 977 MODULE_LICENSE("GPL"); 978